Dear Statalister,
i am running the following 5-class latent class model on a dataset of about 19.000 observations.
gsem (ictcomp_d <-, ologit) (ictapp ictrob itprodimp itperfmon ecommerce <-, logit) (C<- i.country i.est_size i.broad_sector i.est_type), lclass(C 5) from(b) nonrtolerance listwise
everything goes fins till the end of the EM iterations (20), memory consumption at about 40%. As the NR iterations begin the memory consumption jumps to 96% and it fluctuate around that value. I have 16GB RAM, no other applications running.
It is at this point that either Stata crashes or I get the r(3900) message.
I run this model after having run the 2-class, 3-class, and 4-class models and the idea is to get to the 7-class solution.
Notice that a colleague of mine run the same model, from 2-class to a 10-class solution, in Latent Gold without a problem (estimation time less than 10 minutes in the most complex solution), no convergence issue, on a laptop with 8GB RAM and other applications running.
Any idea on what could be going on?
I take this opportunity to wish the list a serene holiday break and a happy new year
Kind regards
Giovanni
0 Response to 5-class latent class model and memory problems r(3900)
Post a Comment