In addition, standard analyses of variance become controversial when you have unbalanced designs, meaning unequal cell sizes. That's fine if you have just a bunch of random subjects, but it creates problems when you have repeated measures or you have independent variables nested within other independent variables. This suggests that observations within the same group should be uncorrelated, or correlated in an unrealistic way. One way to uderstand what is going on, if only vaguely, is to understand that the traditional analyses of variance assumes independent observations, or, with repeated measures, compound symmetry. Just about every source you read on these models takes a somewhat different approach, and it is not always clear how they relate to each other and why they look at the models so differently. This involves focusing on fixed and random models and on repeated measures, and trying to explain why they go together. I am trying to sort out mixed models so that the average reader can understand their purpose and their relationship to one another. This page, or perhaps set of pages, is designed for a different purpose. Some time ago I wrote two web pages on using mixed-models for repeated measures designs. Mixed models Overview Overview of Mixed Models David C.
0 kommentar(er)
