coeftable presents iid se's

[MichelNivard]

When I run this regression:

reg <- feglm(temp ~ 1 + ext_all + ope_all + agr_all + neu_all + con_all | famid, family=binomial ,data=data.final)
reg
GLM estimation, family = binomial, Dep. Var.: temp
Observations: 2,422
Fixed-effects: famid: 1,211
Standard-errors: Clustered (famid) 
         Estimate Std. Error   z value  Pr(>|z|)    
ext_all -0.123273   0.119762 -1.029311 0.3033337    
ope_all  0.301196   0.116957  2.575264 0.0100164 *  
agr_all  0.125027   0.118438  1.055640 0.2911329    
neu_all -0.110300   0.129586 -0.851172 0.3946736    
con_all -0.382102   0.123512 -3.093640 0.0019772 ** 
---

the printed results in inconsistent with coeftable in that coeftable prints the iid standard errors...

> reg$coeftable
          Estimate Std. Error   z value     Pr(>|z|)
ext_all -0.1232727 0.08486547 -1.452566 1.463443e-01
ope_all  0.3011959 0.08201335  3.672523 2.401676e-04
agr_all  0.1250274 0.08344317  1.498354 1.340412e-01
neu_all -0.1103001 0.09104139 -1.211538 2.256893e-01
con_all -0.3821022 0.08661142 -4.411684 1.025698e-05

Which could be desired behaviour I guess, but it'll lead to confusion...

[etiennebacher]

Looks similar to #495