Study
Questions for Quiz #3
1. Be prepared to tell the interpretation of a correlation and a multiple regression weight for each of the following types of variables:
Quantitative variable
Centered quantitative variable
Unit-coded binary variable
Dummy coded binary variable
Effect coded binary variable
Multiple category variable
Dummy code from a set
representing a multiple category variable
Effect code from a set
representing a multiple category variable
Product code from a dummy coded binary and centered quantitative variable
Product code from two centered
quantitative variables
Squared quantitative variable
Centered and squared
quantitative variable
2. Identify the six kinds of regression weights in a 2xQ non-linear interaction model, when each is found, and specifically how to interpret each.
3. What are the advantages of including nonlinear terms in our multiple regression models? How does one decide which to consider? How ought they be examined/tested? What “artifact” must be considered whenever we think we have found a non-linear effect and how to we rule out this artifact?
4. What are the advantages of including multiple-category variables our multiple regression models? How does one decide which to consider? How ought they be included? How ought they be examined/tested?
5. What are the advantages of including interaction terms in our multiple regression models? How does one decide which to consider? How ought they be examined/tested? The presence of an interaction requires what additional careful consideration?
6. What is the difference between the group difference H0: posed by ANOVA, regression, and ANCOVA? What are the internal validity and external validity issues raised when using statistical rather than experimental control of potential confounds? In the examples discussed in class we often found different effects depending upon which covariate we included in the analysis. What concerns does this raise for you? How do you quiet these concerns?
7. Now that you know how to include nonlinear and interaction effects in multivariate models, what are the cautions about doing so (suppressors, collinearity, null flooding)? How should one proceed when considering including these variables (theory & history)?
8. Describe the four (suggested) phases of multivariate statistical analysis, telling the purpose of each and their potential limitations.