Problems with Small STATA

[Ryan Ruddy]

I am teaching an econometrics course and many of my students are using small STATA. In their most recent assignment, I asked students to use the margins command to find marginal effects in probit and logit estimations. It seems that whenever they run margins, stata reports that there is not room to add more variables. They also had a similar problem with random effects models earlier in the semester. I know that small stata has an observations limit limit of 1200 observations and 98 right hand side variables. They are estimating a model with 1000 observations and less than 10 variables. I can trim the data set to meet the constraints, but I don't know what the constraint is.

[ben earnhart]

They're probably hitting limits in the matrix size, try "help limits". Unfortunately, I can't tell you ahead of time how small is small enough, since FE and RE models' matsize depend not only on the observe # cases, but the # of groups.

[Nick Cox]

My guess is different from Ben's. Many programs create temporary variables and that is subject to the limit that is being reported here. Stata doesn't use a rule that variables you only want temporarily don't count. Otherwise that would be a way of subverting the limit.

See e.g. extended previous discussion within http://www.statalist.org/forums/foru...many-variables

and if you want yet more search the forum for "room to add more variables".