Analysis of Categorical Panel Data

s In some categorical tables, one of the classifying variables may be at least ordinal (ran ked) arising fro m a follow-up o r any similar study. The other classifying variab le(s) may be that which separates the population into groups using variables such as gender, race or location, or a co mbination of some of them. The counts obtained this way are analyzed recognizing that one of the variables is nearly metric and must be used and interpretation becomes easier when appropriate model is fitted to the arising product multino mial. An examp le o f such an approach is provided using the data fro m Tuber- culosis Management in a Teaching Hospital. We observed that the recovery rate of females was faster than their males counterpart on the assumption that those discharged through management system follows an exponential distribution.