​Workaround for limits on number of -estimation store- results

[Aaditya Dar]

Hi,

-help limits- reports that no more than 300 estimation results can be stored in Stata. I am running a number of regressions with different permutations and combinations of dependent and independent variables. However, I am unable to hold more than 300 estimates. Some suggestions on this list/stackoverflow have recommended using -statsby- but that doesn't help me (I think) because I am not running the regressions on subsets of the data.

I could save 300 estimates to one .ster file and the remaining estimates to another .ster file but is there is a way to recover estimates from these two separate files using model names/titles (and not number)? Alternatively, could you suggest other ideas to circumvent this upper limit?

Thank you!

[Clyde Schechter]

Why don't you tell us more about where you are trying to go with this. When you use -estimates store- or -estimates save-, you are filling up memory (or your mass storage device) with all of the information in e(), much of which you may never need again. If you are only interested in saving the coefficients and their standard errors, it would make more sense to put your 300+ regressions in a loop and update a -postfile- each time through. If you are trying to save or store the results so that you can later bring them back to run post-estimation commands, then you really do need all the information. In that case, probably your best bet is to choose the names of the .ster files in a way that easily identifies the particular regression model contained in it. The only limit on the number of .ster files you can save is the size of your hard drive.

[Aaditya Dar]

Hi Clyde,

I will have a look at -postfile-. Thank you for the suggestion. I might have to run some post estimation commands but the immediate need is to be able to compare coefficients across different regression models. One quick clarification: there is no restriction on the number of estimates that can be stored in a *single* .ster file, except for limits on the hard disk, right?

[Clyde Schechter]

I do not know of any limits on the number of estimates that can be stored in a single .ster file. But they might exist. I did a quick experiment saving a .ster file with 1000 estimates from some simple linear regressions, and encountered no problems. And I was able to use the estimates from it afterward. So I doubt you will have a problem with 300+..

It dawns on me that apart from the availability of space on the hard drive, the operating system might also impose limits on the size of a file.

[Aaditya Dar]

Thanks, Clyde!