DABB 04 13 Hierarchical Regression Handout
Data Analysis Brown Bag April 2013
Model Building: Hierarchical Regression Data Analysis Brown Bag February 2013
Karen Grace-Martin
©2013 K Grace-Martin
http://TheAnalysisFactor.com
1
Data Analysis Brown Bag April 2013
The Goal Choose variables that: 1. Answers the Research Question 2. Reflects the Real Relationships and Form of the Data
©2013 K Grace-Martin
http://TheAnalysisFactor.com
2
Data Analysis Brown Bag April 2013
The Top-Down Strategy 1. Fit a full model, including proposed interactions 2. Refine the model to account for assumptions, better fit 3. Reduce the model by removing non-significant effects
©2013 K Grace-Martin
http://TheAnalysisFactor.com
3
Data Analysis Brown Bag April 2013
The Step-Up Strategy 1. Start with a basic model 2. Add and test predictors to improve model fit 3. Adjust the final model, if necessary, to account for assumptions
©2013 K Grace-Martin
http://TheAnalysisFactor.com
4
Data Analysis Brown Bag April 2013
Hierarchical Regression 0.1 Choose Variables 0.2 Run Univariate and Bivariate Descriptives, including graphs •
Run a Model with Control Variables
•
Add Key Predictors
•
Add Interactions and/or Curvature
•
Adjust the final model, if necessary, to account for assumptions and data issues
©2013 K Grace-Martin
http://TheAnalysisFactor.com
5
Data Analysis Brown Bag April 2013
Choosing Variables
©2013 K Grace-Martin
http://TheAnalysisFactor.com
6
Data Analysis Brown Bag April 2013
Choosing Variables
©2013 K Grace-Martin
http://TheAnalysisFactor.com
7
Data Analysis Brown Bag April 2013
Choosing Variables Considerations 1. 2. 3. 4.
How well it measures the construct Distribution and missing values Relationship to Y Relationship to other Xs
©2013 K Grace-Martin
http://TheAnalysisFactor.com
8
Data Analysis Brown Bag April 2013
Distribution and Missing Values
©2013 K Grace-Martin
http://TheAnalysisFactor.com
9
Data Analysis Brown Bag April 2013
Distribution and Missing Values
©2013 K Grace-Martin
http://TheAnalysisFactor.com
10
Data Analysis Brown Bag April 2013
Relationship to Y
©2013 K Grace-Martin
http://TheAnalysisFactor.com
11
Data Analysis Brown Bag April 2013
Relationship to Y
©2013 K Grace-Martin
http://TheAnalysisFactor.com
12
Data Analysis Brown Bag April 2013
Relationship to Each Other
©2013 K Grace-Martin
http://TheAnalysisFactor.com
13
Data Analysis Brown Bag April 2013
Relationship to Each Other
©2013 K Grace-Martin
http://TheAnalysisFactor.com
14
Data Analysis Brown Bag April 2013
©2013 K Grace-Martin
http://TheAnalysisFactor.com
15
Data Analysis Brown Bag April 2013
©2013 K Grace-Martin
http://TheAnalysisFactor.com
16
Data Analysis Brown Bag April 2013
Chosen Variables Block 1: Demographic Controls 1. 2. 3.
Marital Status: Married Divorced Separated Widowed MaleDummy NumberBioStepAdoptChildHH2000
Block 2: Psychological and Situational Controls: 1. 2. 3. 4.
CESD2000Total WorkTimeHrsperWkPastCalYr2000cat4 UnemployedWksPastCal2000cont PovertyStatus2000c2
Block 3: Key Predictors 1.
Religious Attendance in 5 Categories
©2013 K Grace-Martin
http://TheAnalysisFactor.com
17
Data Analysis Brown Bag April 2013
Block 1: Demographic Controls R2 = .027 Adj. R2 = .024
©2013 K Grace-Martin
http://TheAnalysisFactor.com
18
Data Analysis Brown Bag April 2013
Block 2: Psychological and Situational Controls R2 = .186 Adj. R2 = .184
Decision: Keep all variables ©2013 K Grace-Martin
http://TheAnalysisFactor.com
19
Data Analysis Brown Bag April 2013
Block 3: Key Predictors R2 = .002 Adj. R2 = .001
R2 = .008 Adj. R2 = .006
Decision: Treat Religious Attendance as Categorical ©2013 K Grace-Martin
http://TheAnalysisFactor.com
20
Data Analysis Brown Bag April 2013
Build Blocks R2 = .025 Adj. R2 = .023
©2013 K Grace-Martin
R2 = .188 Adj. R2 = .183
http://TheAnalysisFactor.com
21
Data Analysis Brown Bag April 2013
Build Blocks R2 = .186 Adj. R2 = .184
©2013 K Grace-Martin
http://TheAnalysisFactor.com
22
Data Analysis Brown Bag April 2013
Build Blocks R2 = .186 Adj. R2 = .184
©2013 K Grace-Martin
http://TheAnalysisFactor.com
R2 = .188 Adj. R2 = .184
23
Data Analysis Brown Bag April 2013
Next Steps 1. Possibly Add Interactions or Quadratic Terms 2. Test Assumptions 3. Check Data Issues
©2013 K Grace-Martin
http://TheAnalysisFactor.com
24
Model Building: Hierarchical Regression Data Analysis Brown Bag February 2013
Karen Grace-Martin
©2013 K Grace-Martin
http://TheAnalysisFactor.com
1
Data Analysis Brown Bag April 2013
The Goal Choose variables that: 1. Answers the Research Question 2. Reflects the Real Relationships and Form of the Data
©2013 K Grace-Martin
http://TheAnalysisFactor.com
2
Data Analysis Brown Bag April 2013
The Top-Down Strategy 1. Fit a full model, including proposed interactions 2. Refine the model to account for assumptions, better fit 3. Reduce the model by removing non-significant effects
©2013 K Grace-Martin
http://TheAnalysisFactor.com
3
Data Analysis Brown Bag April 2013
The Step-Up Strategy 1. Start with a basic model 2. Add and test predictors to improve model fit 3. Adjust the final model, if necessary, to account for assumptions
©2013 K Grace-Martin
http://TheAnalysisFactor.com
4
Data Analysis Brown Bag April 2013
Hierarchical Regression 0.1 Choose Variables 0.2 Run Univariate and Bivariate Descriptives, including graphs •
Run a Model with Control Variables
•
Add Key Predictors
•
Add Interactions and/or Curvature
•
Adjust the final model, if necessary, to account for assumptions and data issues
©2013 K Grace-Martin
http://TheAnalysisFactor.com
5
Data Analysis Brown Bag April 2013
Choosing Variables
©2013 K Grace-Martin
http://TheAnalysisFactor.com
6
Data Analysis Brown Bag April 2013
Choosing Variables
©2013 K Grace-Martin
http://TheAnalysisFactor.com
7
Data Analysis Brown Bag April 2013
Choosing Variables Considerations 1. 2. 3. 4.
How well it measures the construct Distribution and missing values Relationship to Y Relationship to other Xs
©2013 K Grace-Martin
http://TheAnalysisFactor.com
8
Data Analysis Brown Bag April 2013
Distribution and Missing Values
©2013 K Grace-Martin
http://TheAnalysisFactor.com
9
Data Analysis Brown Bag April 2013
Distribution and Missing Values
©2013 K Grace-Martin
http://TheAnalysisFactor.com
10
Data Analysis Brown Bag April 2013
Relationship to Y
©2013 K Grace-Martin
http://TheAnalysisFactor.com
11
Data Analysis Brown Bag April 2013
Relationship to Y
©2013 K Grace-Martin
http://TheAnalysisFactor.com
12
Data Analysis Brown Bag April 2013
Relationship to Each Other
©2013 K Grace-Martin
http://TheAnalysisFactor.com
13
Data Analysis Brown Bag April 2013
Relationship to Each Other
©2013 K Grace-Martin
http://TheAnalysisFactor.com
14
Data Analysis Brown Bag April 2013
©2013 K Grace-Martin
http://TheAnalysisFactor.com
15
Data Analysis Brown Bag April 2013
©2013 K Grace-Martin
http://TheAnalysisFactor.com
16
Data Analysis Brown Bag April 2013
Chosen Variables Block 1: Demographic Controls 1. 2. 3.
Marital Status: Married Divorced Separated Widowed MaleDummy NumberBioStepAdoptChildHH2000
Block 2: Psychological and Situational Controls: 1. 2. 3. 4.
CESD2000Total WorkTimeHrsperWkPastCalYr2000cat4 UnemployedWksPastCal2000cont PovertyStatus2000c2
Block 3: Key Predictors 1.
Religious Attendance in 5 Categories
©2013 K Grace-Martin
http://TheAnalysisFactor.com
17
Data Analysis Brown Bag April 2013
Block 1: Demographic Controls R2 = .027 Adj. R2 = .024
©2013 K Grace-Martin
http://TheAnalysisFactor.com
18
Data Analysis Brown Bag April 2013
Block 2: Psychological and Situational Controls R2 = .186 Adj. R2 = .184
Decision: Keep all variables ©2013 K Grace-Martin
http://TheAnalysisFactor.com
19
Data Analysis Brown Bag April 2013
Block 3: Key Predictors R2 = .002 Adj. R2 = .001
R2 = .008 Adj. R2 = .006
Decision: Treat Religious Attendance as Categorical ©2013 K Grace-Martin
http://TheAnalysisFactor.com
20
Data Analysis Brown Bag April 2013
Build Blocks R2 = .025 Adj. R2 = .023
©2013 K Grace-Martin
R2 = .188 Adj. R2 = .183
http://TheAnalysisFactor.com
21
Data Analysis Brown Bag April 2013
Build Blocks R2 = .186 Adj. R2 = .184
©2013 K Grace-Martin
http://TheAnalysisFactor.com
22
Data Analysis Brown Bag April 2013
Build Blocks R2 = .186 Adj. R2 = .184
©2013 K Grace-Martin
http://TheAnalysisFactor.com
R2 = .188 Adj. R2 = .184
23
Data Analysis Brown Bag April 2013
Next Steps 1. Possibly Add Interactions or Quadratic Terms 2. Test Assumptions 3. Check Data Issues
©2013 K Grace-Martin
http://TheAnalysisFactor.com
24
