[Clyde Schechter]

The -if e(sample)- part is only valid when used after an estimation command (a regression model). So you have to rerun your regression before you use that. Once the regression is run, -if e(sample)- will remain valid until you either run another estimation command (another regression, or -margins, post-, a few others), or close your Stata session.

[Celine Tran]

Clyde Schechter Thank you very much !

[Andrea Smurra]

Probably late, but I had the same problem and found a solution.

TLDR: try adding the following to your code, before running the regression

capture drop __00*

that is, drop any tempvar that might have been stored by mistake.

In my case, I had the same issue: regression runs fine, F-test returns empty.
The same happened if I created new, completely random, continuo variables and ran regressions on these - the issue in my case is not about the the rank of VCV matrix nor similar data issues.

However, the F-test worked if I created new random variables for testing purposes after clearing the dataset, so starting from scratch.
I assume the issue is about the variables in the original dataset - maybe "test" uses a tempvar in its calculation? Seems far fetched, but "fre" would also return an error if there are tempvars stored in the dataset with the name __0000000.
Hopefully the solution above is not a fluke and can help others! It worked for me.