Something I do not quite understand. I am looking into an interaction effect between 2 variables. (UAI & Altman)
I had to winsorize the Altman variable (highonly) in order to get a realistic result (I created the variable Altman_high).
So as follows: I multiplied the UAI (which is on a scale of 1-120) and the WINSORIZED Altman variable (going from -3,85 to 14).
If I take the summary of the interaction variable (Interaction), I notice on first hand there are outliers.
Do I have to winsorize this interaction effect once again? Or can I just use it in my model?
Code:
. gen Interaction= UAI* Altman_high . sum Altman_win Altman_high UAI Interaction Variable | Obs Mean Std. Dev. Min Max -------------+--------------------------------------------------------- Altman_win | 176 5.688867 10.86269 -3.853731 46.44995 Altman_high | 176 3.734336 4.828677 -3.853731 14.24408 UAI | 176 66.88636 22.04952 8 96 Interaction | 176 243.902 299.8308 -327.5671 1310.455
0 Response to Interaction effect outliers?
Post a Comment