Hello!

I am running the same regression on a subsample of my sample as a robustness check. however, on the subsample, the RAMSEY test indicates that there is specification error, while there was no specification error in the original regression. I do know ways to specification errors, but I am doubtful about trying to overcome this specification error. This involves adapting the functional form, adding variables... but then the original regression is not performed anymore and cannot function as a robustness test in my opinion. Are there ways to overcome this problem? Or should I just exclude this robustness test since I cannot interpret it?

Thank you very much!