Hello,

I ran a OLS regression model with 59,000,000 observations and 1000+ plus variables.

When I ran the regression I predicted the fitted values using : predict double chat if e(sample),xb

I also saved the estimates in a .ster file as the file I am using is too big that I don't save it after using the predicted results.

However, when I later used the results from the .ster file, on the same and unchanged dataset, the predictions are slightly different for some observations (at the third and fourth decimal place).

Using the .ster file I just typed: "estimates use my.ster"

I then re-ran the original regression and obtained the predicted values (xb) again, to check if these would be different - and again some observations appear different to both those obtained from the original regression and those from the .ster file.

Is this at all possible - or is it more likely that I am missing something?