Number of groups (e(N_g))

[roberto]

Hello, I am running the following model in Stata.

runmlwin `var' cons Cfem Cage15m psc_a mhedu hhinc5cat Msmk_e_j Malc_e_j Ccp_SDQ11, level2(ks4sch_id: cons) level1(alspid) discrete(d(binomial) l(logit) de(cons) mql1) rigls

although the output (i.e. parameter estimates) seems consistent to when I run the model from within MLwin, the actual number of groups reported in the stata table differ The problem seems to occur when I add in covariates in the fixed part of the model (e.g. psc_a mhedu hhinc5cat Msmk_e_j Malc_e_j Ccp_SDQ11) which may have missing values.

I suggest they check the code that uses and report on the number of groups Any feedback gratefully received.

BW, Roberto.

[GeorgeLeckie]

Thanks Roberto,

Please could you email us the do-file and dataset so we can replicate this problem.

Many thanks

George

[GeorgeLeckie]

Hi Roberto,

The bug you spotted was one that we had actually come across before (it had been spotted by a colleague of yours!). The bug has been fixed in the current release version of runmlwin.

The reason you ran into it was because you had an older version of runmlwin. We are frequently releasing updated versions of runmlwin following all the useful feedback we are getting from you and the other researchers posting messages on this forum.

To determine whether your version of runmlwin is up to date, and to update it if not, issue the following command:

. adoupdate runmlwin

Best wishes

George