I am comparing my model between developed and less developed countries. Some predictors (e.g. CTR in the attached picture) significant in less developed countries model but insignificant in developed countries.I tested whether the difference is significant. The result indicated that the difference is not significant. How can I interpret such result.?
I found an answer in the paper
The Difference Between “Significant” and “Not Significant” is not Itself Statistically Significant
Well, there are several conclusions you can draw- for instance considering CTR it's significant in less developed countries, meaning that there is a strong impact of this variable on the response. For the developed countries this statement is not true, the effect is rather small there, as the coefficient is not significant (it's -0.021, so very small, so one could expect this...). In the full model it's significant again which comes from the less developed countries contained in the full model as well.
Now, considering the difference between the two models we find that CTR doesn't differ significantly between the two models (the difference is -0.103). So basically you have four test results for each variable you are considering, providing you information on the variable with regard to the model (full model/less developed country/developed country) and the difference between the last two.
This is something that you can do in Stata pretty easily using the suest command. See this link for details: https://stats.idre.ucla.edu/stata/code/comparing-regression-coefficients-across-groups-using-suest/
- Badges
- Science topic