What do you do when faced with analyzing student ratings from 1 to 5 for 3 instructors in 3 classes? Aside from questioning the validity of students assessing instructor capability other than for the income generated by high enrollments, and that this is but a toy example, many naive analysts turn to the well-known analysis of variance (ANOVA) using score as a response and instructor and class (eg., time series, linear models, etc.) as in uencing variables to make comparisons among the instructors and classes. While each instructor may teach more than one class, no two instructors share the same class.
This model appears perfectly reasonable, so why do seasoned analysts turn to methods such as those presented in Handbook for Applied Modeling: Non-Gaussian and Correlated Data?
Authors Jamie Riggs and Trent Lalonde explore these models and when they are appropriate.
Read the entire post here
Find out more on their website.