ABSTRACT
Given that we are concerned in this chapter with the situation in which the outcome variable is categorical, that is, longitudinal data with categorical outcome variable, the logistic regression for a binary outcome for example, assumes that observations are independent across time. But it is not uncommon in longitudinal data to at least imagine, for example, if observations were taken at say, time 1, time 2 up to time 6 then observations at times 1 and 2 are more likely to have higher correlations than say, at times 1 and 6.
