Testparm for Interaction terms

[Philipp Gild]

Dear STATALISTers,

I would like to ask your help on an issue I´ve come across in terms of an interaction term, or rather how to interpret:
Specifically, I am running a multivariable logistic regression model (outcome: procedural efficacy of a procedure/surgery) in order to study the impact of a medication (finasterid) on the procedure (prostate enucleation). Overall, finasterid has a positive effect (see A). Next I want to see if the effect of finasterid is modulated by prostate size (variable: pvolquart). In the regression output (see B), the interaction term is significant for the 4th volume quartile of prostate size (which I interpret as: the overall positive impact of the medication is driven by its effect on prostates of the 4th volume quartile, while there is no positive effect in the 1-3rd volume quartile).
However, when I want to confirm the positive interaction via the testparm command (see C), it is no longer significant. Is testparm the incorrect test in this case, or does it mean that there is no significant interaction?

Your help is greatly appreciated!



A)




B)






C)

[Carlo Lazzaro]

Philipp:
-testparm- is simply telling you that, taken jointly, the interaction coefficients cannot reject the null of being=0.
Setting technicalities aside for a while, you're actually incuding two non significant and one significant interaction coefficients: the two non-significant coefficients outperform the significant one.
What above does no imply that the last interaction is not significant (when taken as a single coefficient).
As an aside, in your future posts please use CODE delimiters:

to post what you typed and what Stata gave you back (as per FAQ) and avoid screenshots that are terrible to read and impossible to work on. Thanks

[Philipp Gild]

Thank you for your swift reply and help ! In the future I´ll post the code instead

Kind regards,
Philipp