White and McBurney: Chapters 46 & 14 Chapter 4 Writing in ...
White and McBurney: Chapters 46 & 14 Chapter 4 Writing in psychology •
Most crucial step in research process is communicating results
•
Archival Publication– Written record of scientific process
•
•
•
o
Journals and books
o
The permanent record of science
Invisible Colleges– Informal communication network of people having common scientific interests o
New ideas and results usually discussed through informal communication
o
Allows scientists to keep up with what’s happening in other labs and also permits researchers to present their ideas in a tentative form before committing themselves in archival publications
Discourse Community– A group of people who share common goals, a public forum, common knowledge, and a specialized language o
When writing, must consider discourse community to shape report properly
o
Our discourse community is psychologists, and we expect them to use same style of writing: APA style
The written report o
1
General
The purpose of scientific writing is to convey a message clearly, concisely, and interestingly
Scientific writing aims to convey and inform
Argument– A set of reasons in support of a proposition •
Scientific report is form of this
•
Abstract is overview of argument, introduction gives premise of argument, method tells how evidence was obtained, results presents evidence, discussion draws conclusions from evidence and tells reader where premises came from
Thesis– The proposition that is supported by arguments
A good writing report in 3 words: Clarity, brevity, and felicity
Clarity
•
Most important element of scientific writing is to say exactly what you mean as directly as possible
•
Many words commonly misused in scientific writing (Sternberg, 2003): Affect and effect
Brevity •
Does every word, phrase, or sentence contribute to the paper?
•
Because of space shortage in journals, papers must be as short as possible o
o
Felicity •
How ideas are expressed is important to how well they are accepted
•
Humour is best avoided, but there is a place for wit
•
Write it in a pleasing style so it will get noticed
Avoiding bias in writing
APA reflects recent changes in culture by requiring the avoidance of bias in writing Includes gender, sexual orientation, racial or ethnic identity, disabilities, or age
APA suggests 3 guidelines for avoiding bias:
2
Can sometimes lead to bad style, but can also be aid to good communication
•
Precision in describing people, not man but person
•
Be respectful in labelling people, people with schizophrenia not schizophrenics
•
Acknowledge people who participated in study and avoid passive voice in writing
2 common problems with biased language
o
•
Ambiguity of Referent– When you say man or he instead of them
•
Stereotyping by Labeling– Not all professors are he’s, so have to say them
Avoid labeling people whenever possible
The parts of a paper: Title, author names and affiliations, author note, abstract, introduction, method, results, discussion, references, footnotes, tables with notes, figures with captions
Use computer or word processor, 12point font, times new roman, 1 inch margins, double space, start sections on new page (expect for method, results, and discussion Bolded headings)
Title
•
Gain attention of audience
•
Main idea of paper in a few words
•
Avoid words that don’t directly contribute to paper
Author names and their affiliations •
3
Listed in order of contributions to paper
Author note •
Provides mailing address for reprints, acknowledges financial support or technical assistance, etc.
•
On first page under title, author names, and affiliations
Abstract •
Brief synopsis of paper (150250 words)
•
Summarizes problem, method, results, and conclusion, and contains elements of each major part of paper
•
Most difficult part to write Often written last
Introduction •
Sets stage for rest of paper
•
First, state general problem examined
•
Then discuss relevant literature
4
•
Finally, state how your study will contribute to understanding of problem
•
Indicate hypothesis and expected results
Methods •
Describes what was done in experiment Therefore, in past tense
•
Has 2 purposes: 1. Should explain how to repeat experiment exactly in all essential details; 2. Another person should be able to judge validity of conclusions by comparing them with method section
•
Subsections: Participants How many participants used and how they were obtained, describe characteristics of participants that are important to the study, state that IRB approval was obtained and that APA ethical guidelines were observed in study, name strain, supplier, and species if using animals: Materials Materials used in study, describe stimuli, and give associative values: Design State logic of experiment including the variables, give order that variables were presented, tell what dependent variables were: Procedure Sequence of steps in putting design into effect (doesn’t need to be subsection)
Results •
Indicate data transformations made before analyzing data
•
State what was found
•
Refer reader to table or graph in data
•
Indicate which results significant and what significance tests were used
•
Statistics assure results were not likely a fluke, tables take little space are precise and easy to coordinate with statistical analysis, and figures give better idea of effect size and any interactions Preference for figures
Discussion •
Builds on results by interpreting and relating to literature
•
Focuses on theoretical contribution of study
o
Starts by stating relationship between findings and hypothesis, describe similarities between current results and others, don’t introduce further data, weaknesses/limitations, contribution to understanding problem stated
•
Conclusions found in this section are stated in present tense
References •
Contains the documentation of points made in paper
•
Serves function of trying paper to the literature
Footnotes •
2 types: Content and copyright permission
•
Footnotes to content of paper should be avoided
•
When essential, content footnotes are numbered and appear in journal at bottom of page in which they are cited
•
Copyright permission footnotes are used for reprinted tables of figures
Tables •
Help make data clearer
•
Doesn’t duplicate material in text or figures, but should supplement that material
•
Refer to table in text and explain significance
Figures •
Can describe and refer to figures in text, but they appear last in a type written manuscript
•
Each figure has own page in the manuscript
•
Must be APA format Avoid colour coding, symbols used to indicate groups and conditions, axes clearly labeled, symbols and lettering large enough to be easily read, figures have captions that describe content whis is directly below on same page
Documenting your paper
5
•
Must cite all things not yours unless it is common knowledge
Cite several sources to give reader feel for existing work in area
Plagiarism is presenting other’s work as your own
Paraphrase is when you reword, can’t use three words in a row the same
Unintentional plagiarism occurs when you write something thinking it’s your own words, but you are unconsciously influenced by reading others work Take good notes to prevent this
Selfplagiarism occurs when submitting previously published work as though it were new
Secondary Citation Documentation of an idea from one work that is reported in another one •
Two main techniques for citing other work: 1. Name authours in text followed by year of publication (use “and”), 2. Refer to authours of work and date parenthetically (use “&”)
If 2 authours, always list names; If 3 or more, list all names first time, then only first author with “et al.” and date
Reference has 4 major parts: Authour(s), date, title, and publication information Each separated by period and the rest by commas; Hanging indents
Authours appear in same format for ass references, but titles of books and journals italicized, titles of articles not italicized
Reference to hard copy journal article: Authour(s), date, title, journal (italics), volume (italics), page range, (sometimes issue number) •
Electronic copies same but always need issue number and readers directed to online source
Digital Object Identifiers A unique alphanumeric code that identifies and provides a persistent link to information on the internet
Book: Authour(s), date, title (I), city published, and publisher •
6
If haven’t laid eyes on something, can’t cite work directly
Electronic copies include source location
Article or chapter in book considered edited volume: Authour(s), date, article or chapter title, editor(s), book title (I) and page numbers, publication city, publisher
•
Editor’s names listed forwards
•
Electronic copies same but with DOI or URL
Gray Literature Scientific literature that falls outside the peer review process •
o
•
•
7
When citing, give URL and date retrieved
Steps in the publication process
Choose journal that has published articles on same topic, and choose journal cited most frequently in reference list
Many important steps: Submit over letter, screening and review, editor accepts or conditionally accepts or rejects, revise paper, resubmission with cover letter telling changes made, final decision to accept or reject
If accepted, takes long time (year or more) to be put in upcoming journal This period called in press
Oral presentations o
Same parts as research paper but some changes: Shorter because presented in 1520 minutes, simplify material for audience, don’t just read off paper
o
12 points from each section on presentation outline
o
Introduction in a few sentences, method sticks to essential elements of design, most of time spent on results (emphasize main findings), brief discussion pointing out implications, and summarize results at the end
o
Use visual aids Slides or overhead transparencies
o
Could provide handout if in small group
o
Practice talk beforehand
Poster presentations o
Popular presentation choice in meetings; Ideally suited to class projects; Must be well organized
o
Same information as oral but in static graphic format
o
People can browse and spend time on interesting posters
o
Strip of paper at top with title and authours, abstract in upper left corner on single sheet of paper, rest of material in columns to be read from top bottom and then leftright
o
Use large font, usually 3045 point, tables and figures about 8 by 10”
o
Stand with poster and give information as needed, asked
Chapter 5 Variables •
Variable– Aspect of a testing condition that can change or take on different characteristics with different conditions
•
Reducing a phenomenon to variables focuses researcher’s attention on specific events out of the many that may be related to phenomenon
•
Variables exist in real world but theory is an idea Must make assumptions to relate them
•
Variables are tangible, theoretical concepts are intangible
•
Variables related to theoretical concepts by operational definitions used to measure concepts
•
Relationship between variables provides support for or against the theory that generated experiment
•
Types of variables o
Most basic distinction among variables: Dependent and independent
o
Dependant Variable A measure of the subject’s behaviour that reflects the independent variable’s effects
•
Response of person or animal Could be score on test or behavioural response
•
All of the below are dependant variables
Frequency The number of times that a behaviour is performed •
Rate The number of times that a behaviour is performed relative to time Rate is ratio of frequency to time •
8
Number of times went running
Miles per hour
Duration The amount of time that a behaviour lasts
Latency The amount of time between an instruction and when the behaviour is actually performed
Topography The shape or style of the behaviour •
o
Force– The intensity or strength of a behaviour
Locus Where the behaviour occurs in the environment
Independent Variable– The condition manipulated or selected by the experimenter to determine its effect on behaviour Believed to cause change in dependent variable
The stimulus of stimulusresponse psychology
Stimulus– The cause of something
Levels– The different values of an independent variable •
Variables of Interest– A variable for which its role in the cause and effect of an observed relationship is not clear
o
Independent variable thought of as what experimenter does to subject, and dependent is what subject does back
o
Subject Variables– A difference between subjects that cannot be controlled but can only be selected– A type of independent variable
Confounded Variable– One whose effect cannot be separated from the supposed independent variable
•
When guys like blue and girls like pink, not genetic, but confounded variable because boys were dressed in blue from birth and girls in pink since birth
Not all details of a study are independent variables o
Things that aren’t part of the study are controlled so as to be the same every time the study takes place
o
Only a variable if it changes
Quantitative and categorical variables o
Quantitative Variable– One that varies in amount
9
Poverty, age, IQ for example
Confounded variables o
•
Every independent variable has two values
o
•
Length of stride, measure shape of behaviour
Speed of response
o
Categorical Variable– One that varies in kind
o •
This distinction important for building theories and making graphs
Continuous and discrete variables o
Some quantitative variables can take any value on continuum
o
Continuous Variable– One that falls along a continuum and is not limited to a certain number of values
o
Knowledge in psychology, unlimited
Discrete Variable– One that falls into separate bins with no intermediate values possible
Knowledge in a class, limited, due to test having certain number of questions
o
Although a variable may be continuous, its measurement is often discontinuous
o
Real Limits– The interval defined by the number plus or minus half the distance to the next number
o •
College major or gender
Important for graphing and certain statistical computations
Apparent Limits– The point indicated by a number
Measurement o
The ability to state laws quantitatively is an indication of the scientific progress of that field
o
To state laws quantitatively means two things are true:
The phenomenon is regular enough to make a reasonably precise statement of it •
The law is simple enough to write an equation about it •
•
E=MC squared
What is measurement? o
10
Data that are too variable can obscure any underlying law
Measurement– The process of assigning numbers to events or objects according to rules
•
Properties of the event are represented by properties of the number system
The rules by which numbers are assigned to events determine how useful the measurement is
Types of measurement scales o
o
o
o
11
Nominal Scales– A measure that simply divides objects or events into categories according to their similarities or differences
Simplest scale because its rule of assignment is simplest Objects or events of same kind get same number and different kind get different number
Classification system
Ordinal Scales– A measure that assigns objects or events a name and arranges them in order of their magnitude
Order of vegetables in order of mostliked to mostdisliked
Only order of preference, not how much more or less one is liked over another
Rule: The ordinal position (rank order) of numbers on the scale must represent the rank order of psychological attributes of the objects or events
Interval Scales– A measure in which the differences between numbers are meaningful; Includes both nominal and ordinal information
Rank vegetables from 1 – 7 so can see how much more you like some than others
Rule: Equal differences between numbers on the scale must represent equal psychological differences between events or objects
Ratio Scales– A measure having a meaningful zero point as well as all of the nominal, ordinal, and interval properties
Can use any number that seem appropriate No upper or lower limit
When there is a zero, means indifferent, 10 means you dislike it just as much as you like the vegetable you gave a +10
Rule: The ratios between the numbers on the scale must represent the psychological ratios between the objects of events
•
Comparison of the scales o
As we go from nominal to ordinal to interval to ratio we gain more information
o
Nominal– Two events are same or different
o
Ordinal– That plus, ranking on some variable
o
Interval– That plus, quantitative statements regarding magnitude of differences
o
Ratio– That plus, conveys information about ratios of magnitudes
o
Therefore, we try to make scales of variables ratio scales
o
Permissible Transformations– The ways the assignment of numbers to individual events can be altered without distorting the scale
o
Transformations of data can be performed in order to aid in interpretation or meet assumptions of statistical test, these become fewer as we go from nominal to ratio
o
Ratio scale Can change little without distorting; Can only multiply all the numbers by a positive constant
o
Interval scale Can add or subtract constant from all numbers, or multiply them by positive constant
o
Ordinal Any change that preserves order among members, including both above
o
Nominal Any substitution of a number for another number that preserves similarities and differences, including all above
o
For ratio scale, must be able to be twice something, impossible for IQ
•
Measurement and statistics? o
The scale on which a variable is measured determines the type of statistics that can appropriately be performed on the data
o
Parametric statistics should only be used for interval or ratio– Says psychologists
o
12
120 IQ person isn’t twice as smart as 60 IQ person
Pearson correlation coefficient, the t test, analysis of variance
Nonparametric statistics include various tests that are based on rank order of the data or on the sign of the differences between subjects
•
•
Reliability and validity of measurements o
Reliability– The property of consistency of a measurement that gives the same result on different occasions
o
Validity (of a measurement)– The property of a measurement that tests what it is supposed to test
o
A measurement should be reliable before it can be valid
Variability and error o
•
Error variance– Variability in the dependent variable that is not associated with the independent variable: AKA random error
When changes in the dependent variables are not caused by the independent variable
Bad kind of variability
Validity of measurements o
Construct validity, face validity, content validity, and criterion validity
o
Construct Validity (of a test)– A test that the measurements actually measure the constructs they are designed to measure, but no others
o
o
13
Less frequently used because less power, parametric uses all information present, nonparametric doesn’t
A test of courage should test for only courage, not compassion or anything else
A test with construct validity measures theoretical construct it is supposed to and nothing else, measures what it intends to measure but not any unrelated constructs, and should prove useful in predicting results related to the theoretical construct its measuring
The other kinds of validity are actually different ways of talking about construct validity
Face Validity– Idea that a test should appear superficially (to any person) to test what it is supposed to test
More of a problem of public relations than of true validity
A test can have high or low validity regardless of face validity
Content Validity– Idea that a test should sample the range of behaviour represented by the theoretical concept being tested
o
Should measure everything to do with the construct, if for intelligence, then measure spatial, verbal, and general knowledge, not just one of them
Criterion Validity– Idea that a test should correlate with other measures of the same theoretical construct
Concurrent Validity– If the criterion of the test correlates with something happening currently •
Predictive Validity– If the criterion measures how well someone will do on a future performance •
o •
Types of measurement error: o
2 types: Systematic (or constant) error and random error (or error variance) Variability associated with independent variable isn’t error
o
Systematic Error– Measurement error that is associated with consistent bias
o
Okay if error is there for entire study but if only there for part then results will be skewed
Random Error Variability not associated with any known independent variable
Always problem because it reduces precision with which you assess the effects of independent variable; Threat to reliability of measurement
Types of reliability measures o
TestRetest Reliability– The degree to which the same test score would be obtained on another occasion
o
Internal Consistency– The degree to which the various items on a test are measures of the same thing
14
Intelligence test predicting graduation from college
The concept of correlation is closely intertwined with that of validity
•
Intelligence test correlating with child doing well in school now
Splithalf reliability is a test of internal consistency of dependent variables Items on test divided in half as if they were two different tests, then scores on each half are correlated to see how closely
scores agree on each half; Good tests will have high splithalf correlation
KuderRichardson20 (KR20) computes all possible splithalf correlations on set of data and takes average; Value ranges from 0 (complete disagreement) to 1 (perfect agreement)
Chapter 6 Validity •
Validity– An indication of accuracy in terms of the extent to which a research conclusion corresponds with reality
•
Types of validity o
o
Internal Validity– Extent to which a study provides evidence of a cause effect relationship between the independent and dependent variables
Most fundamental type because it concerns the logic of the relationship between the independent and the dependent variable
In an experiment with high internal validity, it was really the independent variable that caused the dependent variable to change
Confounding– Error that occurs when the effects of two variables in an experiment cannot be separated, resulting in a confused interpretation of the results
o
•
Subject Variables– A difference between subjects that cannot be controlled but can only be selected
Example is gender, cannot be controlled or changed but only selected
Increased chance of confounding
Construct Validity (of research)– Extent to which the results support the theory behind the research o
Can you generalize from specific operations of your experiment to the general theoretical construct about the population in question? Would another theory predict the same experimental results
15
One of the biggest threats to validity in experimentation
Two types of construct validity related Both concerned with how well the underlying idea (or theory) is reflected in the measurement
o
Auxiliary Hypotheses– Other hypotheses that must be true in order to measure your own hypothesis
o
Manipulation Check– Aspect of an experiment designed to make certain that variables have changed in the way that was intended
o
Checks if independent variable is working in the way you thought it was going to
Improves validity of experiment
Example includes the Beck Anxiety Inventory
Construct and internal validity are similar In internal, you strive to rule out alternative variables as potential causes of behaviour of interest; In construct, you must rule out other possible theoretical explanations of the results
•
External Validity– How well the findings of an experiment generalize to other situations or populations (different settings, times, subjects, treatments, observations, etc.) o
People said taboo words more slowly than nontaboo words, in 1949
o
Ecological Validity– Extent to which an experimental situation mimics a realworld situation
•
16
Either way, you need to perform another test to rule out a threat to validity: In internal, redesign study to control for source of confounding; In construct, design new study that permits a choice between two competing theoretical explanations of results
Closely related, but not identical, to external validity
Statistical Conclusion Validity– Extent to which data are shown to be the result of causeeffect relationships rather than accident o
Did dependant variable causes change in independent variable? Or was it random chance?
o
How strong is relationship between independent and dependent variables?
o
To have statistical conclusion validity, must be certain that enough people were tested so that your statistical test can have adequate power, must be sure that the measure of dependent variable is accurate, and must be sure that the inferential statistic chosen for analysis is appropriate for data from independent random samples
o
Even when statistics are used properly, as statistical test only establishes that an outcome has a certain low probability of occurring by chance, does not guarantee that change isn’t result of random error
o
Power– The probability of rejecting the null hypothesis when it is, in fact, false
o •
All method of judging validity only increase confidence in conclusions
Experiments that lack power suffer from lack of statistical conclusion validity
Effect Size– Strength of the relationship between the independent and dependent variables
Threats to validity o
Threats to internal validity
Guarding against threats to internal validity consists of avoiding confounding variables; Major sources of confounding are: Ambiguous temporal precedence, history, maturation, effects of testing, regression effect, selection, and mortality
Ambiguous Temporal Precedence– Although two variables are related, it is not clear which one is that cause and which one is the effect •
Events outside the laboratory (history) •
History– Events that occur outside the experiment that could influence the results of the experiment
Maturation– A source of error in an experiment related to the amount of time between measurements •
Subjects could change between conditions of an experiment due to naturally occurring processes
•
Larger problem with children because they change more rapidly
Effects of testing •
17
Cause must always come before effect, but sometimes unclear
Effects of Repeat Testing– Performance on a second test is influenced by simply having taken a first test
o
Regression Effect– Tendency of subjects with extreme scores on a first measure to score closer to the mean on a second testing •
Operates when there is less than a perfect correlation between variables
•
May occur when two variables are correlated or when same variable is measure twice
•
We often get imperfect correlations when the same variable is measured twice Occurs when error is associated with the measurement of the variable
•
Random Error– That part of the value of a variable that can be attributed to chance
Selection– A confound that can occur due to assignment of subjects to groups •
o
Often not possible to randomly assign individuals to groups because of preexisting groups (e.g., veterans)
Mortality– The dropping out of some subjects before an experiment is completed, causing a threat to validity •
Sometimes called selective subject loss or attrition
•
Biases can be recorded when certain types of people are likely to drop out
Threats to construct validity
Most difficult validity to achieve because of the infinite number of theories that may account for a given lawful relationship
Loose connection between theory and method •
Poor operational definition of theoretical concepts
Ambiguous effect of independent variables •
18
Similar to maturation in that subjects change over time but different in that the change is caused by the testing procedure itself
The participants may perceive the experiment differently than the experimenter The experiment is ambiguous, the independent variable may be manipulated differently in participants
•
o
Threats to external validity
Other subjects •
•
Would the same results be found at another time?
•
Many historical trends render particular research findings invalid (E.g., the dirty word experiment)
Other settings
Threats arise from improper use of statistics when measuring the data
Problem of Power– Too few subjects or too few observations can cause erroneous results
Inaccurate Effect Size Estimation– The size of the relationship is measured poorly Might believe that the independent variable caused a smaller effect than it actually did
The social psychology of the psychology experiment o
Role Demands– Participants’ expectations of what an experimenter requires them to do
o
GoodSubject Tendency– Tendency of experimental participants to act according to what they think the experimenter wants
o
Evaluation Apprehension– Tendency of experimental participants to alter their behaviour to appear as socially desirable as possible
Chapter 14 19
Will the results be the same in another laboratory setting, real world setting?
Threats to statistical conclusion validity
• •
Do the people that you are using represent the type of people you would like to learn about? Participants should be chosen with equal attention to their representativeness relative to some larger population
Other times
• o
The effect of participant expectation can be devastating When aware of being in an experiment, people may act differently Solution is to keep participants from knowing
Data exploration part 1: Graphic and descriptive techniques •
Preparing data for analysis o
Raw Data The sheets on which the subjects’ responses are entered
o
You should decide exactly how you are going to handle data (record it, prepare it for analysis, and what statistics to use) before seeing your first participant
o
Lab Notebook– Notebook for recording information important to a study, such as design, procedure, and the planned analysis
20
Prevents things from being forgotten
o
Once data is collected, it needs to be prepared for data analysis First step called data reduction
o
Components of data handling:
Put data into matrix form in a summary data sheet
Do preliminary statistics and plots
Check for invalid data and make corrections
Check for missing data and replace with missing data code
Check for wild data and remove
Describe data numerically, with descriptive statistics
Describe data graphically
Perform inferential statistics
o
Data Reduction
o
Data Reduction– The process of transcribing data from individual data sheets to a summary form
o
The initial recording of data done on sheets (electronic or hard copy) that record responses of single participant Should have spaces to record time, date, researcher’s name, condition, and other identifying information; Usually participant’s name is in separate place
o
If data collected as questionnaire or other format that participant fills out, place column on righthand side of data sheet into which the data can be transferred as a first step in data reduction Each space in this column should be numbered to correspond to the column in which the data will be recorded in a summary sheet or electronic spread sheet
o
After readying the data the next step is to transcribe the information onto a summary sheet that contains all the data from the study in matrix format •
o •
•
21
Details of data reduction differ but generally put data in matrix form and clearly identify variables important for analysis
The coding guide o
•
Rows subjects, columns are independent variables
Coding Guide– A record that specifies the variables of a study, the columns they occupy in the data file, and their possible values
Necessary for nominal scale to state which label goes with which number
I.e., the 1 = 8 a.m. class, the 2 = 11 a.m. class
Preliminary descriptive statistics o
Descriptive Statistics– Statistics that summarize a set of data
o
Inferential Statistics– Statistics that help us to draw conclusions about populations
o
Before starting inferential statistics we should take a preliminary look at the data, descriptive statistics give an idea as to what the typical score is, how much the scores differ from each other, and can also provide a hint as to errors in the data
o
The most common descriptive statistics are those that concern the average (Measures of Central Tendency– A measure of the average score in a distribution, such as the mean, median or mode) and the variability of a set of data (Measures of Variability– A measure of the degree of difference between scores in a distribution)
Measures of central tendency o
All three are actually kinds of averages Some meanings commonly associated with the term average are: A number that is typical of all scores, a number that is in the middle of all scores, and a number that represents all the scores Each best captures on of them
o
Mode– The most common score in a frequency distribution
Represents most typical score
Stable for large data sets but bounces around for small sets
Does not enter into any further statistical calculations
o
o
•
Not influenced by other scores at all
Median– The middlemost score in a distribution
Requires rank ordering scores from highest to lowest to find middle one N+1/2
Advantage that half scores are above and half below Not affected by how far other scores are from the median, only by how many scores are above and below
Disadvantages: Requires ranking and counting to find middle, limited use for further calculations
Mean– The common average
Adding scores and dividing by number of scores
Advantage: Influenced by every score in the distribution, and mean is the basis for most common and powerful statistical computations, and means of subgroups can be obtained to get mean of entire group (grand mean; median and mode can’t do this)
Disadvantage: Because influenced by every score, it is sensitive to values of extreme scores
Measures of variability o
General types are those based on range, percentiles, and mean
o
Range– Difference between the highest and lowest scores in a distribution Simplest
o
Depends completely on 2 extreme scores, so highly unstable
o
Percentilebased measures
Advantage: Better represent skewed distributions than other measures
Disadvantage: More cumbersome to compute and don’t enter into further statistical calculations
Percentile– A score below which a certain percentage of the cases in a distribution fall; A percentile is a score, not a percentage •
22
Median is 50th percentile
Interquartile Range– Measure of variability defined as the difference between the 75th percentile and the 25th percentile (Q3 – Q1); it is a difference in scores, not percentages
•
Usually used to describe variability of scores around median (although and set of percentiles can be used)
•
25th percentile is the first quartile (Q1), the 75 th percentile is the third quartile (Q3), and so on
•
Includes half the cases in a distribution
•
Not affected by outliers
SemiInterquartile Range– Measure of variability defined as half the interquartile range •
o
Variance and standard deviation
These measures are based on mean
Variance– The average of the squared deviations from the mean
•
Advantage: Enters into other statistical calculations
•
Disadvantage: Not scaled in same units as original scores because its expressed as squared deviations from the mean (convert by getting SD)
Standard Deviation– The square root of the variance; A measure of variability in the same units as the scores being described •
Advantage: Related to other common statistical procedures
•
Other advantages are same as mean: Represents all scores, but affected by outliers
Sum of Squares– The sum of the squared deviations from the mean •
o
• 23
(Q3Q1)/2
Used to compute variance and other statistics
Choice of measure of variability
Range isn’t very useful, depends on two extreme scores
Interquartile and semiinterquartile range are useful in describing data when the median has been used as the measure of central tendency and when data is skewed
Variance and standard deviation are most widely used, they relate to the mean and other common statistics
Tables and graphs
o
Table– A display of data in a matrix format
o
Graph– A representation of data by spatial relationships in a diagram
•
Help summarize data and understand relationships between variables
Usually, xaxis Independent, yaxis Dependent
Tables and graphs of frequency data of one variable o
o
Frequency tables
See how many people earned each score
Rows indicate possible scores on the test, columns indicate particular information about those scores
First column is list of all scores, second column is upper real limits, third column is tally of scores recorded by each students for that mark, fourth column is tally represented by a number
Tables enable a lot of information to be organized and displayed in a clear format
Frequency distributions
Frequency Distribution– A graph that shows the number of scores that fall into specific bins, or divisions of the variable •
Xaxis shows various possible scores, yaxis shows frequency of each score
•
Bars go from half unit below score to half unit above Indicating real limits of scores Bars touch because variable being measured is continuous
Histogram– A frequency distribution in which the frequencies are represented by touching bars •
Makes it easy to find mode
•
Better for less categories
Frequency Polygon– A frequency distribution in which the frequencies are connected by straight lines (line graph in same format as bar graph from above) •
24
Better for more categories because data usually begin to approximate a smooth curve
Normal Curve– A bellshaped curve described by a certain mathematical function •
Describes many frequency distributions that occur in nature
•
Enters into a number of theoretical and practical statistical considerations
Skewed– A distribution that is not symmetrical, longer on one side •
o
o
Behaviour of the mean, median, and mode with various shaped distributions
If symmetrical distribution then mean, median, and mode will be same
If skewed they will be different
Negatively Skewed– Mode > median > mean
Positively Skewed– Opposite^
Mode is not affected by skewedness (not affected by any other scores), median will be pulled in direction of skew (affected by scores above and below), and mean is pulled furthest (affected by distance of scores from middle and by number of scores)
Cumulative Frequency Distributions– A frequency distribution that shows the number of scores that fall at or below a certain score
Can be seen in 5th column of frequency table
Frequency polygons usually used by can use histograms too
Score is represented by its upper real limit •
Includes all scores that fall in or below the entire interval
Always increases as it moves to the right, or stays horizontal (monotonically)
Usually sigmoidal, sshaped (curve is horizontal at either end an steepest in middle •
25
Tail at low end of the distribution (more higher scores) is negatively skewed, tail at high end on the distribution (more lower scores) is positively skewed
If distribution is skewed, steeper part will be toward lower or higher end
o
•
Percentiles
By using a frequency table, you can find the percentiles which are indicators of variability
Using real limits in the graph as guidelines because it determines everyone’s marks that are at or below it due to the addition of half of one more unit, the percentage of people who fall above a certain mark and percentage of people who fall below a certain mark can help to determine important percentiles such as the 75 th percentile by obtaining the marks below the 80th percentile and above the 70th percentile for example (if problematic check page 363)
Tables and graphs that show the relationship between two variables o
o
Scattergrams– A graph showing the responses of a number of individuals on two variables (for a number of individual cases); Visual display of correlational data
I.e., xaxis has scores from test 1, yaxis has scores from test 2 and the point of the graph displays both marks to be seen for each person
No difference between which variable is on the xaxis and yaxis
Often don’t know which variables in scattergram are dependent and independent (unless a variable is manipulated)
Correlation Coefficient– Statistic indicating the strength (and direction) of the relationship between two variables
Correlation and regression
Correlation– The strength of a relationship between two variables (sometimes referred to as the Pearson correlation coefficient or r) •
Can take any value between 1 and +1
•
Correlations less than 0.2 are weak; 0.20.4 are moderately weak; 0.40.6 moderate; 0.60.8 moderately strong; and 0.81.0 strong
•
Pearson correlation coefficient is a measure of a linear function Possible that close relation between variables exist but don’t fit straight line (might fit curved line)
Regression– Predicting the value of one variable from another based on their correlation •
26
If x is this, what is y going to be?
•
Use equation from straight line y’ (predicted) = m(slope)x + b
•
If correlation is 1.0, prediction is perfect; If correlation is 0.0, prediction is impossible
•
The value of m depends on 2 things: The correlation between 2 variables, and the scale of measurement of the variables (the scale on which they are measured is the variability or standard deviation
•
Most of the time scattergrams are not scaled by variability Usually plotted in units in which they are measuredIn this case, slope will not equal r
Variance accounted for •
Squaring the correlation coefficient will determine the variance accounted for in y by x
•
Some proportion of y can be explained by the effect of x
•
When the correlation is 1, we have accounted for all the variability in y when we know x (12 = 1) When correlation is 0, no variability accounted for o
•
Goodness of Fit– The degree to which data match the prediction of a regression line o
o
r2 is sometimes considered a measure of the goodness of fit of the data to the regression line
Tables with one independent and one dependent variable
Dependent variable becomes elements of the matrix, the independent variable becomes the columns •
27
When correlation is inbetween the proportion of variance accounted for is less than the correlation
I.e., columns are 8 a.m. class and 11 a.m. class, rows are mean and SD
Tables become more complicated and informative when there is more than one independent variable and when more information is provided
If more than one independent variable, they would become the rows of the table
o
Graphs of functions (line graphs)
Graph of a function is the most typical graph in psychology
Yaxis represents some dependent variable as a function of an independent variable
Generally, functions that we graph involve a response that varies continuously with changes in the quantitative independent variable
Line Graph– A graphical representation using lines to show relationships between quantitative variables •
o
•
•
28
Bar Graphs– Graphical representation of categorical data in which the heights of separated bars, or columns, show the relationships between variables
Used for categorical data
Bars do not touch, representing different categories of behaviour
Relation between frequency distributions and other graphs o
•
Straight lines usually as it shows empirical data, curved lines need a theory and a mathematical equation that explain it
Frequency distributions are different from other graphs but can be related to them
TimeSeries Graph– A graph in which the abscissa (xaxis) represents time o
Something changing over time
o
Cumulative Record Graph– Shows changes over time in different intervals
Often used for operant conditioning and in behavioural research involving a changing criterion design
Xaxis shows time; Yaxis shows the cumulative number of responses since the beginning of the session
Slope of graph shows rate of responding
Mouse pushed bar enough to get it to top of graph, vertical line back to bottom, then restarts
Indicating variability of the data in a graph
o
Error Bars– In a graph, vertical lines that indicate plus or minus one standard deviation of the data or, less frequently, the standard error of the mean
Vertical line through bar graph bars that indicate the SD of the people who made up the graph •
o
o
BoxandWhisker Plot– A type of graph based on median and percentiles rather than mean and standard deviation
The box extends from the 25th to the 75th percentiles, horizontal line through box indicates median, lower vertical line (or whisker) extends to the 10th percentile, and the upper one to the 90th
Shows clearly how skewed data is, bar graph with error bars does not Therefore, useful for skewed data as well as other situations in which the median is appropriate for the data
Can be used for any graph talked about so far; Not often used in psychology; Can replace frequency distribution since it is based on percentiles
Checking for invalid data, missing data, and outliers
Mistakes can be made in the transcribing process or in other areas, also may want to exclude some wild data
For large data sets it is easier to check for errors in preliminary statistics or graphs from above
Invalid Data– Data points that fall outside the defined range for that variable of data •
Coded number has to be 1 or 2 for 8am or 11am class, if any other number than that then it is invalid
Missing Data Empty cells in a data matrix •
29
i.e., how much guys or girls liked a movie, shorter bar at top of boys meant that most answers were high, long line from top to bottom for girls indicates that some liked it a lot and some not at all
Item skipped or not completed or missed
Outliers– Data points that are highly improbable, although not impossible
•
Could be extreme scores from a normal distribution Don’t want to exclude outliers that come from distribution of interest
•
Potential outliers can come from a different distribution than all the other scores o
Want to exclude these outliers because not truly indicative of distribution
o
Random guesses on a test can be considered because they do not represent the true data
o
Same as not knowing instructions, or giving all the same answers
Most researcher use visual or numerical analyses of data, which are informal methods of spotting outliers to spot outliers •
One criterion might be to eliminate responses that fall 3 or more SD away from the mean of the data
•
Necessary to report the data that was eliminated and to describe criteria used to do so
•
Wise to perform statistical analysis with and without wild data and report difference in results
Most simple and systematic way of searching for invalid data, missing data, and outliers is to do some preliminary statistics on the matrix
Missing Data Code– A symbol, such as an asterisk (*), that is entered in a cell that has no data •
Computer will ignore it and adjusts the number of cases accordingly
ChartJunk– Parts of a graph that aren’t necessary to understand it
Lecture Notes: Communication • •
30
Communication important for 2 reasons: Pollination and improving efficiency by avoiding wasted effort Pollination We make progress faster when we test our ideas against other people’s understanding of a problem– The more people the better o An important characteristic of science is that it is international Need to talk to people all over world
Science brings together people of different cultures, having different perspectives on most nonscientific questions. Avoiding wasted effort Communication increases efficiency o Anything we can do to avoid wasted effort attracts more money to research Writing is selfpresentation You will be judged on the basis of what you write o Publishing leads to: Credit for your work, access to the invisible college, and jobs, research grants, and influence o Neatness indicates care Suggests project was worth you effort o
• •
Lecture Notes: Variables • • •
•
Variables are things whose values: (a) can be measured, and (b) vary from one occasion to another The dimensions of interest may be physical or abstract o Can never measure an abstract dimension itself What would we measure? We can measure proxy variables– For example, by creating a questionnaire or other instrument o A problem: The relation of scores on the instrument to scores on the underlying dimension is very often unknown Dependent variables: o Reaction Time Interval between presentation of a stimulus and onset of a response, typically measured in msec o Accuracy Proportion correct in a classification or naming response o Likert Scale A format for attitude measures in which subjects express agreement or disagreement with statements, on a scale from strongly disagree to strongly agree Disagreement on whether its an interval, ordinal, or nominal scale o Test score Measures performance on a test of ability, achievement, or a psychological trait or state o Rate Number of observations per unit of time, as in # of problems solved per hour or # of widgets made per day o Duration How long some state or activity lasts (e.g., duration of vigilance in an attention task) o Physiological Measures Such as galvanic skin response or heart rate (e.g., Hugo Critchley’s work on emotion & interoception) o Independent variable is when we are sure about causeeffect relationship, variables of interest when unsure o Scientists not interested in things that don’t change Interested in variability o Systematic variation can be explained, random variation can’t
Lecture Notes: Tables & Graphs • • • 31
Advantage of using tables: Effect sizes can be calculated Disadvantage of using tables: Patterns in data are more difficult to see in tables than in graphs Most graphs are twodimensional, using a Cartesian coordinate system (X and Y)
• • •
•
•
Each point on the line is an (X, Y) pair – made from one X score and one Y score for each case Bar Graphs: Height indicates D.V.; Location along X axis indicates I.V. Can we objectively measure how deceptive a graph is? Yes, if we can agree on an operationalization– Such as the Graph Discrepancy Index (GDI) o GDI = (a / b – 1) X 100% o a = Percentage change in centimeters shown in graph– That is, [height of high column– Height of low column]; b = Percentage change in the data o Positive values mean that the graph exaggerates the trend; Negative values mean that the graph understates the trend Line Graphs: Point show actual data, lines connecting points show interpolations o Use when response varies continuously wit I.V. o Interpolation Inferring the Y value at an X between two known X values o Extrapolation Inferring the Y value at an X beyond the range of X values for which you have data In linear functions, a unit change in X is always associated with a unit change in Y
Lecture Notes: Validity of an Argument • • • •
•
•
•
32
Validity can be threatened either by poor theory or by poor measurement o We won’t see the expected relationship if: The theory is incorrect, the operationalizations are flawed Internal Validity Is the relationship between variables causal? Construct Validity Does the independent variable capture our theoretical cause? Does the dependent variable capture our theoretical effect? Tactics used to increase internal validity include various kinds of control These controls make your testing situation more specific and thus less typical o As a result, things you do to increase internal validity may decrease external validity and vicevera Poor Reliability: o Increase number of items used in the task o Use more precise measuring instruments o Decrease error by working in a lab, where you have more control Threats to Statistical Conclusion Validity: o Weak relationship between IV and DV Increase strength by increasing “dose”; Decrease “noise” by increasing reliability or decreasing distractions in the environment Threats to Construct Validity: o MonoOperation Bias There may be several ways to operationalize your treatment o MonoMethod Bias Issue is similar here, but concerns testing instrument o Interaction of Different Treatments Might the combination of treatments be the effective treatment?
Most crucial step in research process is communicating results
•
Archival Publication– Written record of scientific process
•
•
•
o
Journals and books
o
The permanent record of science
Invisible Colleges– Informal communication network of people having common scientific interests o
New ideas and results usually discussed through informal communication
o
Allows scientists to keep up with what’s happening in other labs and also permits researchers to present their ideas in a tentative form before committing themselves in archival publications
Discourse Community– A group of people who share common goals, a public forum, common knowledge, and a specialized language o
When writing, must consider discourse community to shape report properly
o
Our discourse community is psychologists, and we expect them to use same style of writing: APA style
The written report o
1
General
The purpose of scientific writing is to convey a message clearly, concisely, and interestingly
Scientific writing aims to convey and inform
Argument– A set of reasons in support of a proposition •
Scientific report is form of this
•
Abstract is overview of argument, introduction gives premise of argument, method tells how evidence was obtained, results presents evidence, discussion draws conclusions from evidence and tells reader where premises came from
Thesis– The proposition that is supported by arguments
A good writing report in 3 words: Clarity, brevity, and felicity
Clarity
•
Most important element of scientific writing is to say exactly what you mean as directly as possible
•
Many words commonly misused in scientific writing (Sternberg, 2003): Affect and effect
Brevity •
Does every word, phrase, or sentence contribute to the paper?
•
Because of space shortage in journals, papers must be as short as possible o
o
Felicity •
How ideas are expressed is important to how well they are accepted
•
Humour is best avoided, but there is a place for wit
•
Write it in a pleasing style so it will get noticed
Avoiding bias in writing
APA reflects recent changes in culture by requiring the avoidance of bias in writing Includes gender, sexual orientation, racial or ethnic identity, disabilities, or age
APA suggests 3 guidelines for avoiding bias:
2
Can sometimes lead to bad style, but can also be aid to good communication
•
Precision in describing people, not man but person
•
Be respectful in labelling people, people with schizophrenia not schizophrenics
•
Acknowledge people who participated in study and avoid passive voice in writing
2 common problems with biased language
o
•
Ambiguity of Referent– When you say man or he instead of them
•
Stereotyping by Labeling– Not all professors are he’s, so have to say them
Avoid labeling people whenever possible
The parts of a paper: Title, author names and affiliations, author note, abstract, introduction, method, results, discussion, references, footnotes, tables with notes, figures with captions
Use computer or word processor, 12point font, times new roman, 1 inch margins, double space, start sections on new page (expect for method, results, and discussion Bolded headings)
Title
•
Gain attention of audience
•
Main idea of paper in a few words
•
Avoid words that don’t directly contribute to paper
Author names and their affiliations •
3
Listed in order of contributions to paper
Author note •
Provides mailing address for reprints, acknowledges financial support or technical assistance, etc.
•
On first page under title, author names, and affiliations
Abstract •
Brief synopsis of paper (150250 words)
•
Summarizes problem, method, results, and conclusion, and contains elements of each major part of paper
•
Most difficult part to write Often written last
Introduction •
Sets stage for rest of paper
•
First, state general problem examined
•
Then discuss relevant literature
4
•
Finally, state how your study will contribute to understanding of problem
•
Indicate hypothesis and expected results
Methods •
Describes what was done in experiment Therefore, in past tense
•
Has 2 purposes: 1. Should explain how to repeat experiment exactly in all essential details; 2. Another person should be able to judge validity of conclusions by comparing them with method section
•
Subsections: Participants How many participants used and how they were obtained, describe characteristics of participants that are important to the study, state that IRB approval was obtained and that APA ethical guidelines were observed in study, name strain, supplier, and species if using animals: Materials Materials used in study, describe stimuli, and give associative values: Design State logic of experiment including the variables, give order that variables were presented, tell what dependent variables were: Procedure Sequence of steps in putting design into effect (doesn’t need to be subsection)
Results •
Indicate data transformations made before analyzing data
•
State what was found
•
Refer reader to table or graph in data
•
Indicate which results significant and what significance tests were used
•
Statistics assure results were not likely a fluke, tables take little space are precise and easy to coordinate with statistical analysis, and figures give better idea of effect size and any interactions Preference for figures
Discussion •
Builds on results by interpreting and relating to literature
•
Focuses on theoretical contribution of study
o
Starts by stating relationship between findings and hypothesis, describe similarities between current results and others, don’t introduce further data, weaknesses/limitations, contribution to understanding problem stated
•
Conclusions found in this section are stated in present tense
References •
Contains the documentation of points made in paper
•
Serves function of trying paper to the literature
Footnotes •
2 types: Content and copyright permission
•
Footnotes to content of paper should be avoided
•
When essential, content footnotes are numbered and appear in journal at bottom of page in which they are cited
•
Copyright permission footnotes are used for reprinted tables of figures
Tables •
Help make data clearer
•
Doesn’t duplicate material in text or figures, but should supplement that material
•
Refer to table in text and explain significance
Figures •
Can describe and refer to figures in text, but they appear last in a type written manuscript
•
Each figure has own page in the manuscript
•
Must be APA format Avoid colour coding, symbols used to indicate groups and conditions, axes clearly labeled, symbols and lettering large enough to be easily read, figures have captions that describe content whis is directly below on same page
Documenting your paper
5
•
Must cite all things not yours unless it is common knowledge
Cite several sources to give reader feel for existing work in area
Plagiarism is presenting other’s work as your own
Paraphrase is when you reword, can’t use three words in a row the same
Unintentional plagiarism occurs when you write something thinking it’s your own words, but you are unconsciously influenced by reading others work Take good notes to prevent this
Selfplagiarism occurs when submitting previously published work as though it were new
Secondary Citation Documentation of an idea from one work that is reported in another one •
Two main techniques for citing other work: 1. Name authours in text followed by year of publication (use “and”), 2. Refer to authours of work and date parenthetically (use “&”)
If 2 authours, always list names; If 3 or more, list all names first time, then only first author with “et al.” and date
Reference has 4 major parts: Authour(s), date, title, and publication information Each separated by period and the rest by commas; Hanging indents
Authours appear in same format for ass references, but titles of books and journals italicized, titles of articles not italicized
Reference to hard copy journal article: Authour(s), date, title, journal (italics), volume (italics), page range, (sometimes issue number) •
Electronic copies same but always need issue number and readers directed to online source
Digital Object Identifiers A unique alphanumeric code that identifies and provides a persistent link to information on the internet
Book: Authour(s), date, title (I), city published, and publisher •
6
If haven’t laid eyes on something, can’t cite work directly
Electronic copies include source location
Article or chapter in book considered edited volume: Authour(s), date, article or chapter title, editor(s), book title (I) and page numbers, publication city, publisher
•
Editor’s names listed forwards
•
Electronic copies same but with DOI or URL
Gray Literature Scientific literature that falls outside the peer review process •
o
•
•
7
When citing, give URL and date retrieved
Steps in the publication process
Choose journal that has published articles on same topic, and choose journal cited most frequently in reference list
Many important steps: Submit over letter, screening and review, editor accepts or conditionally accepts or rejects, revise paper, resubmission with cover letter telling changes made, final decision to accept or reject
If accepted, takes long time (year or more) to be put in upcoming journal This period called in press
Oral presentations o
Same parts as research paper but some changes: Shorter because presented in 1520 minutes, simplify material for audience, don’t just read off paper
o
12 points from each section on presentation outline
o
Introduction in a few sentences, method sticks to essential elements of design, most of time spent on results (emphasize main findings), brief discussion pointing out implications, and summarize results at the end
o
Use visual aids Slides or overhead transparencies
o
Could provide handout if in small group
o
Practice talk beforehand
Poster presentations o
Popular presentation choice in meetings; Ideally suited to class projects; Must be well organized
o
Same information as oral but in static graphic format
o
People can browse and spend time on interesting posters
o
Strip of paper at top with title and authours, abstract in upper left corner on single sheet of paper, rest of material in columns to be read from top bottom and then leftright
o
Use large font, usually 3045 point, tables and figures about 8 by 10”
o
Stand with poster and give information as needed, asked
Chapter 5 Variables •
Variable– Aspect of a testing condition that can change or take on different characteristics with different conditions
•
Reducing a phenomenon to variables focuses researcher’s attention on specific events out of the many that may be related to phenomenon
•
Variables exist in real world but theory is an idea Must make assumptions to relate them
•
Variables are tangible, theoretical concepts are intangible
•
Variables related to theoretical concepts by operational definitions used to measure concepts
•
Relationship between variables provides support for or against the theory that generated experiment
•
Types of variables o
Most basic distinction among variables: Dependent and independent
o
Dependant Variable A measure of the subject’s behaviour that reflects the independent variable’s effects
•
Response of person or animal Could be score on test or behavioural response
•
All of the below are dependant variables
Frequency The number of times that a behaviour is performed •
Rate The number of times that a behaviour is performed relative to time Rate is ratio of frequency to time •
8
Number of times went running
Miles per hour
Duration The amount of time that a behaviour lasts
Latency The amount of time between an instruction and when the behaviour is actually performed
Topography The shape or style of the behaviour •
o
Force– The intensity or strength of a behaviour
Locus Where the behaviour occurs in the environment
Independent Variable– The condition manipulated or selected by the experimenter to determine its effect on behaviour Believed to cause change in dependent variable
The stimulus of stimulusresponse psychology
Stimulus– The cause of something
Levels– The different values of an independent variable •
Variables of Interest– A variable for which its role in the cause and effect of an observed relationship is not clear
o
Independent variable thought of as what experimenter does to subject, and dependent is what subject does back
o
Subject Variables– A difference between subjects that cannot be controlled but can only be selected– A type of independent variable
Confounded Variable– One whose effect cannot be separated from the supposed independent variable
•
When guys like blue and girls like pink, not genetic, but confounded variable because boys were dressed in blue from birth and girls in pink since birth
Not all details of a study are independent variables o
Things that aren’t part of the study are controlled so as to be the same every time the study takes place
o
Only a variable if it changes
Quantitative and categorical variables o
Quantitative Variable– One that varies in amount
9
Poverty, age, IQ for example
Confounded variables o
•
Every independent variable has two values
o
•
Length of stride, measure shape of behaviour
Speed of response
o
Categorical Variable– One that varies in kind
o •
This distinction important for building theories and making graphs
Continuous and discrete variables o
Some quantitative variables can take any value on continuum
o
Continuous Variable– One that falls along a continuum and is not limited to a certain number of values
o
Knowledge in psychology, unlimited
Discrete Variable– One that falls into separate bins with no intermediate values possible
Knowledge in a class, limited, due to test having certain number of questions
o
Although a variable may be continuous, its measurement is often discontinuous
o
Real Limits– The interval defined by the number plus or minus half the distance to the next number
o •
College major or gender
Important for graphing and certain statistical computations
Apparent Limits– The point indicated by a number
Measurement o
The ability to state laws quantitatively is an indication of the scientific progress of that field
o
To state laws quantitatively means two things are true:
The phenomenon is regular enough to make a reasonably precise statement of it •
The law is simple enough to write an equation about it •
•
E=MC squared
What is measurement? o
10
Data that are too variable can obscure any underlying law
Measurement– The process of assigning numbers to events or objects according to rules
•
Properties of the event are represented by properties of the number system
The rules by which numbers are assigned to events determine how useful the measurement is
Types of measurement scales o
o
o
o
11
Nominal Scales– A measure that simply divides objects or events into categories according to their similarities or differences
Simplest scale because its rule of assignment is simplest Objects or events of same kind get same number and different kind get different number
Classification system
Ordinal Scales– A measure that assigns objects or events a name and arranges them in order of their magnitude
Order of vegetables in order of mostliked to mostdisliked
Only order of preference, not how much more or less one is liked over another
Rule: The ordinal position (rank order) of numbers on the scale must represent the rank order of psychological attributes of the objects or events
Interval Scales– A measure in which the differences between numbers are meaningful; Includes both nominal and ordinal information
Rank vegetables from 1 – 7 so can see how much more you like some than others
Rule: Equal differences between numbers on the scale must represent equal psychological differences between events or objects
Ratio Scales– A measure having a meaningful zero point as well as all of the nominal, ordinal, and interval properties
Can use any number that seem appropriate No upper or lower limit
When there is a zero, means indifferent, 10 means you dislike it just as much as you like the vegetable you gave a +10
Rule: The ratios between the numbers on the scale must represent the psychological ratios between the objects of events
•
Comparison of the scales o
As we go from nominal to ordinal to interval to ratio we gain more information
o
Nominal– Two events are same or different
o
Ordinal– That plus, ranking on some variable
o
Interval– That plus, quantitative statements regarding magnitude of differences
o
Ratio– That plus, conveys information about ratios of magnitudes
o
Therefore, we try to make scales of variables ratio scales
o
Permissible Transformations– The ways the assignment of numbers to individual events can be altered without distorting the scale
o
Transformations of data can be performed in order to aid in interpretation or meet assumptions of statistical test, these become fewer as we go from nominal to ratio
o
Ratio scale Can change little without distorting; Can only multiply all the numbers by a positive constant
o
Interval scale Can add or subtract constant from all numbers, or multiply them by positive constant
o
Ordinal Any change that preserves order among members, including both above
o
Nominal Any substitution of a number for another number that preserves similarities and differences, including all above
o
For ratio scale, must be able to be twice something, impossible for IQ
•
Measurement and statistics? o
The scale on which a variable is measured determines the type of statistics that can appropriately be performed on the data
o
Parametric statistics should only be used for interval or ratio– Says psychologists
o
12
120 IQ person isn’t twice as smart as 60 IQ person
Pearson correlation coefficient, the t test, analysis of variance
Nonparametric statistics include various tests that are based on rank order of the data or on the sign of the differences between subjects
•
•
Reliability and validity of measurements o
Reliability– The property of consistency of a measurement that gives the same result on different occasions
o
Validity (of a measurement)– The property of a measurement that tests what it is supposed to test
o
A measurement should be reliable before it can be valid
Variability and error o
•
Error variance– Variability in the dependent variable that is not associated with the independent variable: AKA random error
When changes in the dependent variables are not caused by the independent variable
Bad kind of variability
Validity of measurements o
Construct validity, face validity, content validity, and criterion validity
o
Construct Validity (of a test)– A test that the measurements actually measure the constructs they are designed to measure, but no others
o
o
13
Less frequently used because less power, parametric uses all information present, nonparametric doesn’t
A test of courage should test for only courage, not compassion or anything else
A test with construct validity measures theoretical construct it is supposed to and nothing else, measures what it intends to measure but not any unrelated constructs, and should prove useful in predicting results related to the theoretical construct its measuring
The other kinds of validity are actually different ways of talking about construct validity
Face Validity– Idea that a test should appear superficially (to any person) to test what it is supposed to test
More of a problem of public relations than of true validity
A test can have high or low validity regardless of face validity
Content Validity– Idea that a test should sample the range of behaviour represented by the theoretical concept being tested
o
Should measure everything to do with the construct, if for intelligence, then measure spatial, verbal, and general knowledge, not just one of them
Criterion Validity– Idea that a test should correlate with other measures of the same theoretical construct
Concurrent Validity– If the criterion of the test correlates with something happening currently •
Predictive Validity– If the criterion measures how well someone will do on a future performance •
o •
Types of measurement error: o
2 types: Systematic (or constant) error and random error (or error variance) Variability associated with independent variable isn’t error
o
Systematic Error– Measurement error that is associated with consistent bias
o
Okay if error is there for entire study but if only there for part then results will be skewed
Random Error Variability not associated with any known independent variable
Always problem because it reduces precision with which you assess the effects of independent variable; Threat to reliability of measurement
Types of reliability measures o
TestRetest Reliability– The degree to which the same test score would be obtained on another occasion
o
Internal Consistency– The degree to which the various items on a test are measures of the same thing
14
Intelligence test predicting graduation from college
The concept of correlation is closely intertwined with that of validity
•
Intelligence test correlating with child doing well in school now
Splithalf reliability is a test of internal consistency of dependent variables Items on test divided in half as if they were two different tests, then scores on each half are correlated to see how closely
scores agree on each half; Good tests will have high splithalf correlation
KuderRichardson20 (KR20) computes all possible splithalf correlations on set of data and takes average; Value ranges from 0 (complete disagreement) to 1 (perfect agreement)
Chapter 6 Validity •
Validity– An indication of accuracy in terms of the extent to which a research conclusion corresponds with reality
•
Types of validity o
o
Internal Validity– Extent to which a study provides evidence of a cause effect relationship between the independent and dependent variables
Most fundamental type because it concerns the logic of the relationship between the independent and the dependent variable
In an experiment with high internal validity, it was really the independent variable that caused the dependent variable to change
Confounding– Error that occurs when the effects of two variables in an experiment cannot be separated, resulting in a confused interpretation of the results
o
•
Subject Variables– A difference between subjects that cannot be controlled but can only be selected
Example is gender, cannot be controlled or changed but only selected
Increased chance of confounding
Construct Validity (of research)– Extent to which the results support the theory behind the research o
Can you generalize from specific operations of your experiment to the general theoretical construct about the population in question? Would another theory predict the same experimental results
15
One of the biggest threats to validity in experimentation
Two types of construct validity related Both concerned with how well the underlying idea (or theory) is reflected in the measurement
o
Auxiliary Hypotheses– Other hypotheses that must be true in order to measure your own hypothesis
o
Manipulation Check– Aspect of an experiment designed to make certain that variables have changed in the way that was intended
o
Checks if independent variable is working in the way you thought it was going to
Improves validity of experiment
Example includes the Beck Anxiety Inventory
Construct and internal validity are similar In internal, you strive to rule out alternative variables as potential causes of behaviour of interest; In construct, you must rule out other possible theoretical explanations of the results
•
External Validity– How well the findings of an experiment generalize to other situations or populations (different settings, times, subjects, treatments, observations, etc.) o
People said taboo words more slowly than nontaboo words, in 1949
o
Ecological Validity– Extent to which an experimental situation mimics a realworld situation
•
16
Either way, you need to perform another test to rule out a threat to validity: In internal, redesign study to control for source of confounding; In construct, design new study that permits a choice between two competing theoretical explanations of results
Closely related, but not identical, to external validity
Statistical Conclusion Validity– Extent to which data are shown to be the result of causeeffect relationships rather than accident o
Did dependant variable causes change in independent variable? Or was it random chance?
o
How strong is relationship between independent and dependent variables?
o
To have statistical conclusion validity, must be certain that enough people were tested so that your statistical test can have adequate power, must be sure that the measure of dependent variable is accurate, and must be sure that the inferential statistic chosen for analysis is appropriate for data from independent random samples
o
Even when statistics are used properly, as statistical test only establishes that an outcome has a certain low probability of occurring by chance, does not guarantee that change isn’t result of random error
o
Power– The probability of rejecting the null hypothesis when it is, in fact, false
o •
All method of judging validity only increase confidence in conclusions
Experiments that lack power suffer from lack of statistical conclusion validity
Effect Size– Strength of the relationship between the independent and dependent variables
Threats to validity o
Threats to internal validity
Guarding against threats to internal validity consists of avoiding confounding variables; Major sources of confounding are: Ambiguous temporal precedence, history, maturation, effects of testing, regression effect, selection, and mortality
Ambiguous Temporal Precedence– Although two variables are related, it is not clear which one is that cause and which one is the effect •
Events outside the laboratory (history) •
History– Events that occur outside the experiment that could influence the results of the experiment
Maturation– A source of error in an experiment related to the amount of time between measurements •
Subjects could change between conditions of an experiment due to naturally occurring processes
•
Larger problem with children because they change more rapidly
Effects of testing •
17
Cause must always come before effect, but sometimes unclear
Effects of Repeat Testing– Performance on a second test is influenced by simply having taken a first test
o
Regression Effect– Tendency of subjects with extreme scores on a first measure to score closer to the mean on a second testing •
Operates when there is less than a perfect correlation between variables
•
May occur when two variables are correlated or when same variable is measure twice
•
We often get imperfect correlations when the same variable is measured twice Occurs when error is associated with the measurement of the variable
•
Random Error– That part of the value of a variable that can be attributed to chance
Selection– A confound that can occur due to assignment of subjects to groups •
o
Often not possible to randomly assign individuals to groups because of preexisting groups (e.g., veterans)
Mortality– The dropping out of some subjects before an experiment is completed, causing a threat to validity •
Sometimes called selective subject loss or attrition
•
Biases can be recorded when certain types of people are likely to drop out
Threats to construct validity
Most difficult validity to achieve because of the infinite number of theories that may account for a given lawful relationship
Loose connection between theory and method •
Poor operational definition of theoretical concepts
Ambiguous effect of independent variables •
18
Similar to maturation in that subjects change over time but different in that the change is caused by the testing procedure itself
The participants may perceive the experiment differently than the experimenter The experiment is ambiguous, the independent variable may be manipulated differently in participants
•
o
Threats to external validity
Other subjects •
•
Would the same results be found at another time?
•
Many historical trends render particular research findings invalid (E.g., the dirty word experiment)
Other settings
Threats arise from improper use of statistics when measuring the data
Problem of Power– Too few subjects or too few observations can cause erroneous results
Inaccurate Effect Size Estimation– The size of the relationship is measured poorly Might believe that the independent variable caused a smaller effect than it actually did
The social psychology of the psychology experiment o
Role Demands– Participants’ expectations of what an experimenter requires them to do
o
GoodSubject Tendency– Tendency of experimental participants to act according to what they think the experimenter wants
o
Evaluation Apprehension– Tendency of experimental participants to alter their behaviour to appear as socially desirable as possible
Chapter 14 19
Will the results be the same in another laboratory setting, real world setting?
Threats to statistical conclusion validity
• •
Do the people that you are using represent the type of people you would like to learn about? Participants should be chosen with equal attention to their representativeness relative to some larger population
Other times
• o
The effect of participant expectation can be devastating When aware of being in an experiment, people may act differently Solution is to keep participants from knowing
Data exploration part 1: Graphic and descriptive techniques •
Preparing data for analysis o
Raw Data The sheets on which the subjects’ responses are entered
o
You should decide exactly how you are going to handle data (record it, prepare it for analysis, and what statistics to use) before seeing your first participant
o
Lab Notebook– Notebook for recording information important to a study, such as design, procedure, and the planned analysis
20
Prevents things from being forgotten
o
Once data is collected, it needs to be prepared for data analysis First step called data reduction
o
Components of data handling:
Put data into matrix form in a summary data sheet
Do preliminary statistics and plots
Check for invalid data and make corrections
Check for missing data and replace with missing data code
Check for wild data and remove
Describe data numerically, with descriptive statistics
Describe data graphically
Perform inferential statistics
o
Data Reduction
o
Data Reduction– The process of transcribing data from individual data sheets to a summary form
o
The initial recording of data done on sheets (electronic or hard copy) that record responses of single participant Should have spaces to record time, date, researcher’s name, condition, and other identifying information; Usually participant’s name is in separate place
o
If data collected as questionnaire or other format that participant fills out, place column on righthand side of data sheet into which the data can be transferred as a first step in data reduction Each space in this column should be numbered to correspond to the column in which the data will be recorded in a summary sheet or electronic spread sheet
o
After readying the data the next step is to transcribe the information onto a summary sheet that contains all the data from the study in matrix format •
o •
•
21
Details of data reduction differ but generally put data in matrix form and clearly identify variables important for analysis
The coding guide o
•
Rows subjects, columns are independent variables
Coding Guide– A record that specifies the variables of a study, the columns they occupy in the data file, and their possible values
Necessary for nominal scale to state which label goes with which number
I.e., the 1 = 8 a.m. class, the 2 = 11 a.m. class
Preliminary descriptive statistics o
Descriptive Statistics– Statistics that summarize a set of data
o
Inferential Statistics– Statistics that help us to draw conclusions about populations
o
Before starting inferential statistics we should take a preliminary look at the data, descriptive statistics give an idea as to what the typical score is, how much the scores differ from each other, and can also provide a hint as to errors in the data
o
The most common descriptive statistics are those that concern the average (Measures of Central Tendency– A measure of the average score in a distribution, such as the mean, median or mode) and the variability of a set of data (Measures of Variability– A measure of the degree of difference between scores in a distribution)
Measures of central tendency o
All three are actually kinds of averages Some meanings commonly associated with the term average are: A number that is typical of all scores, a number that is in the middle of all scores, and a number that represents all the scores Each best captures on of them
o
Mode– The most common score in a frequency distribution
Represents most typical score
Stable for large data sets but bounces around for small sets
Does not enter into any further statistical calculations
o
o
•
Not influenced by other scores at all
Median– The middlemost score in a distribution
Requires rank ordering scores from highest to lowest to find middle one N+1/2
Advantage that half scores are above and half below Not affected by how far other scores are from the median, only by how many scores are above and below
Disadvantages: Requires ranking and counting to find middle, limited use for further calculations
Mean– The common average
Adding scores and dividing by number of scores
Advantage: Influenced by every score in the distribution, and mean is the basis for most common and powerful statistical computations, and means of subgroups can be obtained to get mean of entire group (grand mean; median and mode can’t do this)
Disadvantage: Because influenced by every score, it is sensitive to values of extreme scores
Measures of variability o
General types are those based on range, percentiles, and mean
o
Range– Difference between the highest and lowest scores in a distribution Simplest
o
Depends completely on 2 extreme scores, so highly unstable
o
Percentilebased measures
Advantage: Better represent skewed distributions than other measures
Disadvantage: More cumbersome to compute and don’t enter into further statistical calculations
Percentile– A score below which a certain percentage of the cases in a distribution fall; A percentile is a score, not a percentage •
22
Median is 50th percentile
Interquartile Range– Measure of variability defined as the difference between the 75th percentile and the 25th percentile (Q3 – Q1); it is a difference in scores, not percentages
•
Usually used to describe variability of scores around median (although and set of percentiles can be used)
•
25th percentile is the first quartile (Q1), the 75 th percentile is the third quartile (Q3), and so on
•
Includes half the cases in a distribution
•
Not affected by outliers
SemiInterquartile Range– Measure of variability defined as half the interquartile range •
o
Variance and standard deviation
These measures are based on mean
Variance– The average of the squared deviations from the mean
•
Advantage: Enters into other statistical calculations
•
Disadvantage: Not scaled in same units as original scores because its expressed as squared deviations from the mean (convert by getting SD)
Standard Deviation– The square root of the variance; A measure of variability in the same units as the scores being described •
Advantage: Related to other common statistical procedures
•
Other advantages are same as mean: Represents all scores, but affected by outliers
Sum of Squares– The sum of the squared deviations from the mean •
o
• 23
(Q3Q1)/2
Used to compute variance and other statistics
Choice of measure of variability
Range isn’t very useful, depends on two extreme scores
Interquartile and semiinterquartile range are useful in describing data when the median has been used as the measure of central tendency and when data is skewed
Variance and standard deviation are most widely used, they relate to the mean and other common statistics
Tables and graphs
o
Table– A display of data in a matrix format
o
Graph– A representation of data by spatial relationships in a diagram
•
Help summarize data and understand relationships between variables
Usually, xaxis Independent, yaxis Dependent
Tables and graphs of frequency data of one variable o
o
Frequency tables
See how many people earned each score
Rows indicate possible scores on the test, columns indicate particular information about those scores
First column is list of all scores, second column is upper real limits, third column is tally of scores recorded by each students for that mark, fourth column is tally represented by a number
Tables enable a lot of information to be organized and displayed in a clear format
Frequency distributions
Frequency Distribution– A graph that shows the number of scores that fall into specific bins, or divisions of the variable •
Xaxis shows various possible scores, yaxis shows frequency of each score
•
Bars go from half unit below score to half unit above Indicating real limits of scores Bars touch because variable being measured is continuous
Histogram– A frequency distribution in which the frequencies are represented by touching bars •
Makes it easy to find mode
•
Better for less categories
Frequency Polygon– A frequency distribution in which the frequencies are connected by straight lines (line graph in same format as bar graph from above) •
24
Better for more categories because data usually begin to approximate a smooth curve
Normal Curve– A bellshaped curve described by a certain mathematical function •
Describes many frequency distributions that occur in nature
•
Enters into a number of theoretical and practical statistical considerations
Skewed– A distribution that is not symmetrical, longer on one side •
o
o
Behaviour of the mean, median, and mode with various shaped distributions
If symmetrical distribution then mean, median, and mode will be same
If skewed they will be different
Negatively Skewed– Mode > median > mean
Positively Skewed– Opposite^
Mode is not affected by skewedness (not affected by any other scores), median will be pulled in direction of skew (affected by scores above and below), and mean is pulled furthest (affected by distance of scores from middle and by number of scores)
Cumulative Frequency Distributions– A frequency distribution that shows the number of scores that fall at or below a certain score
Can be seen in 5th column of frequency table
Frequency polygons usually used by can use histograms too
Score is represented by its upper real limit •
Includes all scores that fall in or below the entire interval
Always increases as it moves to the right, or stays horizontal (monotonically)
Usually sigmoidal, sshaped (curve is horizontal at either end an steepest in middle •
25
Tail at low end of the distribution (more higher scores) is negatively skewed, tail at high end on the distribution (more lower scores) is positively skewed
If distribution is skewed, steeper part will be toward lower or higher end
o
•
Percentiles
By using a frequency table, you can find the percentiles which are indicators of variability
Using real limits in the graph as guidelines because it determines everyone’s marks that are at or below it due to the addition of half of one more unit, the percentage of people who fall above a certain mark and percentage of people who fall below a certain mark can help to determine important percentiles such as the 75 th percentile by obtaining the marks below the 80th percentile and above the 70th percentile for example (if problematic check page 363)
Tables and graphs that show the relationship between two variables o
o
Scattergrams– A graph showing the responses of a number of individuals on two variables (for a number of individual cases); Visual display of correlational data
I.e., xaxis has scores from test 1, yaxis has scores from test 2 and the point of the graph displays both marks to be seen for each person
No difference between which variable is on the xaxis and yaxis
Often don’t know which variables in scattergram are dependent and independent (unless a variable is manipulated)
Correlation Coefficient– Statistic indicating the strength (and direction) of the relationship between two variables
Correlation and regression
Correlation– The strength of a relationship between two variables (sometimes referred to as the Pearson correlation coefficient or r) •
Can take any value between 1 and +1
•
Correlations less than 0.2 are weak; 0.20.4 are moderately weak; 0.40.6 moderate; 0.60.8 moderately strong; and 0.81.0 strong
•
Pearson correlation coefficient is a measure of a linear function Possible that close relation between variables exist but don’t fit straight line (might fit curved line)
Regression– Predicting the value of one variable from another based on their correlation •
26
If x is this, what is y going to be?
•
Use equation from straight line y’ (predicted) = m(slope)x + b
•
If correlation is 1.0, prediction is perfect; If correlation is 0.0, prediction is impossible
•
The value of m depends on 2 things: The correlation between 2 variables, and the scale of measurement of the variables (the scale on which they are measured is the variability or standard deviation
•
Most of the time scattergrams are not scaled by variability Usually plotted in units in which they are measuredIn this case, slope will not equal r
Variance accounted for •
Squaring the correlation coefficient will determine the variance accounted for in y by x
•
Some proportion of y can be explained by the effect of x
•
When the correlation is 1, we have accounted for all the variability in y when we know x (12 = 1) When correlation is 0, no variability accounted for o
•
Goodness of Fit– The degree to which data match the prediction of a regression line o
o
r2 is sometimes considered a measure of the goodness of fit of the data to the regression line
Tables with one independent and one dependent variable
Dependent variable becomes elements of the matrix, the independent variable becomes the columns •
27
When correlation is inbetween the proportion of variance accounted for is less than the correlation
I.e., columns are 8 a.m. class and 11 a.m. class, rows are mean and SD
Tables become more complicated and informative when there is more than one independent variable and when more information is provided
If more than one independent variable, they would become the rows of the table
o
Graphs of functions (line graphs)
Graph of a function is the most typical graph in psychology
Yaxis represents some dependent variable as a function of an independent variable
Generally, functions that we graph involve a response that varies continuously with changes in the quantitative independent variable
Line Graph– A graphical representation using lines to show relationships between quantitative variables •
o
•
•
28
Bar Graphs– Graphical representation of categorical data in which the heights of separated bars, or columns, show the relationships between variables
Used for categorical data
Bars do not touch, representing different categories of behaviour
Relation between frequency distributions and other graphs o
•
Straight lines usually as it shows empirical data, curved lines need a theory and a mathematical equation that explain it
Frequency distributions are different from other graphs but can be related to them
TimeSeries Graph– A graph in which the abscissa (xaxis) represents time o
Something changing over time
o
Cumulative Record Graph– Shows changes over time in different intervals
Often used for operant conditioning and in behavioural research involving a changing criterion design
Xaxis shows time; Yaxis shows the cumulative number of responses since the beginning of the session
Slope of graph shows rate of responding
Mouse pushed bar enough to get it to top of graph, vertical line back to bottom, then restarts
Indicating variability of the data in a graph
o
Error Bars– In a graph, vertical lines that indicate plus or minus one standard deviation of the data or, less frequently, the standard error of the mean
Vertical line through bar graph bars that indicate the SD of the people who made up the graph •
o
o
BoxandWhisker Plot– A type of graph based on median and percentiles rather than mean and standard deviation
The box extends from the 25th to the 75th percentiles, horizontal line through box indicates median, lower vertical line (or whisker) extends to the 10th percentile, and the upper one to the 90th
Shows clearly how skewed data is, bar graph with error bars does not Therefore, useful for skewed data as well as other situations in which the median is appropriate for the data
Can be used for any graph talked about so far; Not often used in psychology; Can replace frequency distribution since it is based on percentiles
Checking for invalid data, missing data, and outliers
Mistakes can be made in the transcribing process or in other areas, also may want to exclude some wild data
For large data sets it is easier to check for errors in preliminary statistics or graphs from above
Invalid Data– Data points that fall outside the defined range for that variable of data •
Coded number has to be 1 or 2 for 8am or 11am class, if any other number than that then it is invalid
Missing Data Empty cells in a data matrix •
29
i.e., how much guys or girls liked a movie, shorter bar at top of boys meant that most answers were high, long line from top to bottom for girls indicates that some liked it a lot and some not at all
Item skipped or not completed or missed
Outliers– Data points that are highly improbable, although not impossible
•
Could be extreme scores from a normal distribution Don’t want to exclude outliers that come from distribution of interest
•
Potential outliers can come from a different distribution than all the other scores o
Want to exclude these outliers because not truly indicative of distribution
o
Random guesses on a test can be considered because they do not represent the true data
o
Same as not knowing instructions, or giving all the same answers
Most researcher use visual or numerical analyses of data, which are informal methods of spotting outliers to spot outliers •
One criterion might be to eliminate responses that fall 3 or more SD away from the mean of the data
•
Necessary to report the data that was eliminated and to describe criteria used to do so
•
Wise to perform statistical analysis with and without wild data and report difference in results
Most simple and systematic way of searching for invalid data, missing data, and outliers is to do some preliminary statistics on the matrix
Missing Data Code– A symbol, such as an asterisk (*), that is entered in a cell that has no data •
Computer will ignore it and adjusts the number of cases accordingly
ChartJunk– Parts of a graph that aren’t necessary to understand it
Lecture Notes: Communication • •
30
Communication important for 2 reasons: Pollination and improving efficiency by avoiding wasted effort Pollination We make progress faster when we test our ideas against other people’s understanding of a problem– The more people the better o An important characteristic of science is that it is international Need to talk to people all over world
Science brings together people of different cultures, having different perspectives on most nonscientific questions. Avoiding wasted effort Communication increases efficiency o Anything we can do to avoid wasted effort attracts more money to research Writing is selfpresentation You will be judged on the basis of what you write o Publishing leads to: Credit for your work, access to the invisible college, and jobs, research grants, and influence o Neatness indicates care Suggests project was worth you effort o
• •
Lecture Notes: Variables • • •
•
Variables are things whose values: (a) can be measured, and (b) vary from one occasion to another The dimensions of interest may be physical or abstract o Can never measure an abstract dimension itself What would we measure? We can measure proxy variables– For example, by creating a questionnaire or other instrument o A problem: The relation of scores on the instrument to scores on the underlying dimension is very often unknown Dependent variables: o Reaction Time Interval between presentation of a stimulus and onset of a response, typically measured in msec o Accuracy Proportion correct in a classification or naming response o Likert Scale A format for attitude measures in which subjects express agreement or disagreement with statements, on a scale from strongly disagree to strongly agree Disagreement on whether its an interval, ordinal, or nominal scale o Test score Measures performance on a test of ability, achievement, or a psychological trait or state o Rate Number of observations per unit of time, as in # of problems solved per hour or # of widgets made per day o Duration How long some state or activity lasts (e.g., duration of vigilance in an attention task) o Physiological Measures Such as galvanic skin response or heart rate (e.g., Hugo Critchley’s work on emotion & interoception) o Independent variable is when we are sure about causeeffect relationship, variables of interest when unsure o Scientists not interested in things that don’t change Interested in variability o Systematic variation can be explained, random variation can’t
Lecture Notes: Tables & Graphs • • • 31
Advantage of using tables: Effect sizes can be calculated Disadvantage of using tables: Patterns in data are more difficult to see in tables than in graphs Most graphs are twodimensional, using a Cartesian coordinate system (X and Y)
• • •
•
•
Each point on the line is an (X, Y) pair – made from one X score and one Y score for each case Bar Graphs: Height indicates D.V.; Location along X axis indicates I.V. Can we objectively measure how deceptive a graph is? Yes, if we can agree on an operationalization– Such as the Graph Discrepancy Index (GDI) o GDI = (a / b – 1) X 100% o a = Percentage change in centimeters shown in graph– That is, [height of high column– Height of low column]; b = Percentage change in the data o Positive values mean that the graph exaggerates the trend; Negative values mean that the graph understates the trend Line Graphs: Point show actual data, lines connecting points show interpolations o Use when response varies continuously wit I.V. o Interpolation Inferring the Y value at an X between two known X values o Extrapolation Inferring the Y value at an X beyond the range of X values for which you have data In linear functions, a unit change in X is always associated with a unit change in Y
Lecture Notes: Validity of an Argument • • • •
•
•
•
32
Validity can be threatened either by poor theory or by poor measurement o We won’t see the expected relationship if: The theory is incorrect, the operationalizations are flawed Internal Validity Is the relationship between variables causal? Construct Validity Does the independent variable capture our theoretical cause? Does the dependent variable capture our theoretical effect? Tactics used to increase internal validity include various kinds of control These controls make your testing situation more specific and thus less typical o As a result, things you do to increase internal validity may decrease external validity and vicevera Poor Reliability: o Increase number of items used in the task o Use more precise measuring instruments o Decrease error by working in a lab, where you have more control Threats to Statistical Conclusion Validity: o Weak relationship between IV and DV Increase strength by increasing “dose”; Decrease “noise” by increasing reliability or decreasing distractions in the environment Threats to Construct Validity: o MonoOperation Bias There may be several ways to operationalize your treatment o MonoMethod Bias Issue is similar here, but concerns testing instrument o Interaction of Different Treatments Might the combination of treatments be the effective treatment?
