When running STATA for panel data (especially for did_imputation),I was suggested by the code sometimes to increase the number of tolerance and iteration, I am wondering how these things affect my regression result in general ?