Go back over the conceptual questions listed on some of the homeworks. And in addition look at:
Chapter 14: 1, 2, 3, 4, 5 and 6
All material from previous quizzes is examinable. In particular be prepared to interpret coefficients based on full model specifications (i.e. β form) and R output.
When is it valid to exclude an outlying observation from a data analysis?
Know how to identify from a simple response versus explanatory plot, whether an outlier will have large leverage, cook’s distance and/or studentized residual?
What property of an outlying observation does each case influence statistic attempt to capture?
Describe how a two-way analysis of variance is a special case of multiple linear regression.
What are the shorthand forms for the additive and saturated model in a two way ANOVA setting? How many parameters are needed for each, and how many degrees of freedom are associated with each?
How can one choose between the additive model and the saturated model?
Describe the relationship between the effects of the two factors in the additive model? How does that differ in the saturated model?
Display (13.21 in Sleuth and slides from Feb 12), shows some hypothetical treatment curves. You should be able to identify from a plot like these, whether the appropriate model is additive or nonadditive.
Be prepared to interpret F-tests between additive and saturated models. This make take the form of R output, or the Sleuth style ANOVA tables.
Be prepared to interpret individual parameter estimates in a two-way ANOVA setting, either from an additive or saturated model.
What is a replicate?
Given a study description you should be able to identify what the experimental unit is, and how many replicates of each treatment combination there are.
Why is having no replicates a bad thing?
Describe two strategies to deal with the problem of not having enough degrees of freedom to estimate the saturated model. How can you tell if either is appropriate/innappropriate?
When might you want to keep replicates low?