Increase in Adj. R2

Hi everyone,

I have a short question: For my master theses, I am running a simple regression model with several independent variables to examine their impact on the dependent variable. My supervisor also told my that I should include my ind. variables step-wise and to consider the increase in adj. R2 in order to evaluate their importance for explaining the variation of my dep. variable.

The problem is that I have a ind. variable for which the adj R2 is 0.0000 if only this variable is included (and the coefficient is insignificant).
However, when the other ind. variables are already included and I then include this variable, the adj R2 increases by 0.0351 (and the coefficient is significant).

Now I'm a little bit confused. Depending on the order of inclusion this variable has either a relatively strong impact on the increase in adj R2 or it has non effect at all. Have I made a mistake or what is a possible interpreation here?

Thanks a lot for your help!


Not a robit
This is why stepwise is total **** and should not be used. There are countless examples of why stepwise is ****. Variable selection should be based on your content knowledge not a golem. Put the variables your theory supports and just use the partial R^2 values with confidence intervals to understand each variable's effect on DV variability.

P.S., If you are putting multiple IVs into a model you are not doing simple regression, but multiple regression. Simple is when you only have one IV.
Cheers and tell your supervisor they are wrong in their suggestion.