I want to estimate a multilevel logit model. But I'm confused about the minimum number of groups and observations per group. What would be the minimum number of observations per group? My case: I have a small sample over 200 observations that I can classify in 40 groups, but there are some groups with only one observation. It's possible to estimate the model with this approach? Would be better classify in 5 groups with more observations per group?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
There are no rules for the minimum number of groups, when the number of groups is small, it is difficult to estimate the between-group variation and, as a result, multilevel modeling often adds little in such situations (Gelman & Hill, 2007). If we have only two groups classical form is prefered. Two observations per groups is enough to fit a multilevel model.