Study Questions for Quiz #4
(Parenthetical portions will not appear on the exam, but should be used to select and organize information for you’re
your answers)
Classification & Linear Discriminant Function Analysis
1. Compare and contrast prediction vs. classification and describe how the F-test relates to classification and classification accuracy.
2. Describe the different ways to determine if an ldf model “works” and the different ways to describe “how well it works”. (Be sure to include a description of the "% variance" output from SPSS.)
3. Describe the ldf variate. Tell three different bases for "interpreting" one and the relative advantages/disadvantages of each -- what should you do?
4. Describe the occurrence of “suppressor effects”, “multivariate power”, “null washout” and “extreme collinearity” in ldf analyses and any ways that these differ from the occurrence in multiple regression.
5. Distinguish between “concentrated” and “diffuse” ldf structures. Discuss this distinction in terms of "profile differences" among groups (the pattern of between group mean differences across the variables), as well as in terms of classification accuracy.
6. Describe the alternative types of “follow-up analyses” when doing an ldf analysis? Compare and contrast the major types of follow-up analyses in terms of the information derived from each and how to select which type to use for a particular analysis.
Cluster Analysis
7. Describe cluster analysis and compare how and when it is used to linear discriminant function analysis and factor analysis.
8. Describe the findings that might prompt a cluster analysis and what the researcher hopes to accomplish using this technique. Give an example of how cluster analysis might be applied in your research area.
9. Tell the various techniques to help select the “right” number of clusters.
10. Describe the uses of ANOVA and discriminant analysis as a follow-up analysis when clustering.
Factor Analysis
11. Describe the usual application and basic steps in a factor analysis, including how these differ for PC and PAF analyses. Tell how a researcher is likely to choose between PC and PAF models.
12. Compare and contrast the "world view" and purposes of principal components and common factor models. What is the major procedural difference between PC and PAF models. Given this, when can we expect these models to be most similar & most different?
13. Describe and tell the advantages and disadvantages of the major approaches to determining the number of factors (including Bartlett’s test of the correlation matrix, Bartlett’s test of residuals, λ > 1.00, scree, replication, meaningfulness). Why, when using, SPSS are the number of PCs and PAFs always the same?
14. What is meant by factor extraction and how does the purpose of this mathematical operation differ from our usual purpose in conducting a factor analysis. What are the consequences of this difference and how does factor rotation help given our usual purpose?
15. What does it mean to “interpret” a factor solution? What are the usual cutoffs applied during this process, what are the consequences of changing these cutoffs and how does one decide which to select?
16. Tell the difference between why variables load together on a factor and why they load together on an ldf. What are multivocal variables, how are they similar and different in a factoring and ldf, and how do we approach them using each model?