xtlogit error

[Zhexiao Lin]

when i use xtlogit, i encounter error

error: 4,585 (group size) take 1,738 (# positives) combinations results in numeric overflow; computations cannot proceed

[Joseph Coveney]

It seems that you're using xtlogit , fe and that T = 4585. It's difficult to guess what you're doing, but did you accidentally use an indicator factor variable notation (i.) on a continuous predictor?

According to the user's manual for it, "clogit uses a recursive algorithm to compute the likelihood, which means that there are no limits on the size of T_i", and so you might want to look into that alternative, although both commands might be using the same underlying estimation code.

Finally, from the same source, "If T_i is large for all groups, the bias of the unconditional fixed-effects estimator is not a concern, and we can confidently use logit with an indicator variable for each group (provided, of course, that the number of groups does not exceed matsize", and so you might be able to fall back on that if you really do intend to have such T for each group.