Hi All,
I have a balanced panel dataset - 100 companies tracked over 5 year period. My aim is to test hypothesis for certain variables that affect presence/absence of chief innovation officer. Therefore I am using xtgee family (binomial) link (logit). However, I am having problem with deciding on the correlation structure. From my understanding, since the same companies are tracked, correlation is expected. That's why I initially used exchangeable structure. However, for none of the regressions I ran did the p value come out below .1. On the other hand, when I use independent correlation for similar regressions, all come out to be highly significant - p<0.001. Can someone please guide me what I might be missing here?
Thanks a lot!
Mohsin
I have a balanced panel dataset - 100 companies tracked over 5 year period. My aim is to test hypothesis for certain variables that affect presence/absence of chief innovation officer. Therefore I am using xtgee family (binomial) link (logit). However, I am having problem with deciding on the correlation structure. From my understanding, since the same companies are tracked, correlation is expected. That's why I initially used exchangeable structure. However, for none of the regressions I ran did the p value come out below .1. On the other hand, when I use independent correlation for similar regressions, all come out to be highly significant - p<0.001. Can someone please guide me what I might be missing here?
Thanks a lot!
Mohsin