[Carlo Lazzaro]

Marcel:
run .xtlogit,re- with -boostrap- standard errors, which gives the -lr- outcome back; if the null of -rho=0- is rejected, stick with -xtlogit,re-; otherwise you can choose between -xtlogit,pa- and pooled -logit-

[Marcel Issa]

Carlo: Since, I cannot reject the LR test and t

he pseudo r squared does not apply to the xtlogit

what should I make my decision between xtlogit and pooled logit based on?

[Carlo Lazzaro]

Marcel:
since you cannot reject the null (ie, no evidence of random effect detected as per your data), you should go pooled -logit-.

[Marcel Issa]

Carlo:
Is there any way to get the explanatory power or goodness of fit of a xtlogit model?

[Carlo Lazzaro]

Marcel:
see -estat ic- (please note that -xtlogit,pa- does not support this test);
- see -help linktest-.