At the moment I am writing my Thesis. I want to test the moderating effect of industry relatedness on a divestiture - firm performance relationship using subsample analysis.
However, when testing the moderation effect of industry relatedness (1 if related divestiture, 0 if not), then the non divesting firms (IDV "divestiture" = 0) in the sample will not take on a meaningful value since they did not divest in the first place. Thus, for those companies there is no point in assessing whether relatedness influences their performance, since industry relatedness only regards firms that did divest. How should i tackle this?
Hi Jeroen,
Testing a moderation effect with a dichotomous variable (divestiture), can be handled a few ways.
1) you could handle the modelling as a group comparison, depending on how many you have in your sample. Thus you will have two models; one for divesting firms and those who did not. You will get different parameters and you will need to test if they truly are different (e.g., we are talking margin of error, Wald test, etc. depending upon your methodology).
2) you might model an actual moderation effect on one element of your model (i.e., regression parameter). The result of the parameter, while ignoring the moderation effect is for non-divesting firms (i.e., divesting = 0; no additional effect). The result of the regression parameter + the moderating value is for divesting firms (i.e., divesting = 1).
I'm guessing you are doing regression, or path analysis. If you methodology is using SEM and has a measurement model then you will need to deal with testing for invariance. However, if this is not the case ... this will sound like gobbly-gook, and you can ignore!!
Here's hoping for a simple solution, and good results.
Regards,
Tim
If firm performance is a quantitative variable, it sounds like a straightforward two-way ANOVA model. If you Google with the name of the software you use, you should find lots of examples. It may be that the simple main effect of X (divestiture) is significant at one level of M (relatedness) but not at the other. But that does not directly test whether the simple main effects of X differ. Use the F-test on the interaction term to address that question. HTH.
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Writing