Dear Statalisters,
I encountered following issue which I'm not sure how to deal with. So I'm trying to estimate a family fixed effects model with a sets of smaller in size fixed effects for location and schooling. I also want to interact these smaller sets of fixed effects with dummy variable (say marriage). So I do either xtreg or areg where I specify maternal fixed effects as the main FE variable and then I just create sets of dummies for the other ones (not using i. but actually creating it by hand). Thus I have ~3000 school effects/dummies (times two due to marriage interactions) ~ 800 location effects/dummies (times two due to marriage interactions) and some other auxiliary variables. In total I have about 8000 variables in the model plus the absorbed mother FE. I set the matsize to 11000 so to my mind this should handle the inversion of 8K - by - 8K matrix (my understanding is that maternal FE do not count here because data are first transformed before running the actual regression). However, I get the well-known message "matsize too small"; "You have attempted to create a matrix with too many rows or columns or attempted to fit a model with too many variables...". I'm trying felsdvreg as an alternative to take out the set of 3000 school effects from the main matrix inversion but I'd really love to understand what is happening and if I can run it using areg or xtreg (the felsdvreg has been running for some time now). It seems also that this is about Stata and not computer capacity as I still have 150 GB RAM free.
Thanks a ton for your help,
Chris
I encountered following issue which I'm not sure how to deal with. So I'm trying to estimate a family fixed effects model with a sets of smaller in size fixed effects for location and schooling. I also want to interact these smaller sets of fixed effects with dummy variable (say marriage). So I do either xtreg or areg where I specify maternal fixed effects as the main FE variable and then I just create sets of dummies for the other ones (not using i. but actually creating it by hand). Thus I have ~3000 school effects/dummies (times two due to marriage interactions) ~ 800 location effects/dummies (times two due to marriage interactions) and some other auxiliary variables. In total I have about 8000 variables in the model plus the absorbed mother FE. I set the matsize to 11000 so to my mind this should handle the inversion of 8K - by - 8K matrix (my understanding is that maternal FE do not count here because data are first transformed before running the actual regression). However, I get the well-known message "matsize too small"; "You have attempted to create a matrix with too many rows or columns or attempted to fit a model with too many variables...". I'm trying felsdvreg as an alternative to take out the set of 3000 school effects from the main matrix inversion but I'd really love to understand what is happening and if I can run it using areg or xtreg (the felsdvreg has been running for some time now). It seems also that this is about Stata and not computer capacity as I still have 150 GB RAM free.
Thanks a ton for your help,
Chris
