Dear all,
I have two relatively small Data sets (N=76 and N=111) and I use a WLS approach with aweights using the standard errors, since it is a meta regression. I follow the UCLA regression diagnostics (https://stats.idre.ucla.edu/stata/we...n-diagnostics/) and nearly everything works fine.
In the end, -ovtest- shows a p-value of 0,0139 which indicates a rejection of H0. Additionaly, I have some leverage points that exceed the value that is calculated as by UCLA: (2k+2)/n with k=#predictors. My value is 0.24, the highest leverage points are arround 0.70. Can the result of -ovtest- be due to those leverage points or the sample size? As far as I know, the -ovtest- should not have a problem with small sample sizes, but maybe there is something that I did not see. Is such a result by -ovtest- a big problem? Unfortunately, I do not have the biggest experience in applied econometrics. I would be very grateful for any remarks.
Best regards
Janik
I have two relatively small Data sets (N=76 and N=111) and I use a WLS approach with aweights using the standard errors, since it is a meta regression. I follow the UCLA regression diagnostics (https://stats.idre.ucla.edu/stata/we...n-diagnostics/) and nearly everything works fine.
In the end, -ovtest- shows a p-value of 0,0139 which indicates a rejection of H0. Additionaly, I have some leverage points that exceed the value that is calculated as by UCLA: (2k+2)/n with k=#predictors. My value is 0.24, the highest leverage points are arround 0.70. Can the result of -ovtest- be due to those leverage points or the sample size? As far as I know, the -ovtest- should not have a problem with small sample sizes, but maybe there is something that I did not see. Is such a result by -ovtest- a big problem? Unfortunately, I do not have the biggest experience in applied econometrics. I would be very grateful for any remarks.
Best regards
Janik
