Interactions

It is now time to add interaction to our model. My existing variables are exclusiveness, number of reviews, and ratings, so I could try to add a combination of these three. I could also try to add interactions with variables that were not significant when I was working through backwards stepwise selection process.

So I will create the following 2 interaction terms to try to add to my model. If this were a real analysis I would have tried any and all combinations to be sure that I could create the most value possible. I am starting with the dataset where outliers have been removed.

And now I will divide my dataset into the training set and the test set, using the training set to build my model.

It seems that both interactions are not significant and therefore do not contribute to R-squared. In fact it actually made Exclusiveness to no longer be significant. The interesting piece is that after exclusiveness is no longer accounted for, our R-squared increased. In my previous post we saw that R-squared was 49.26% and it has now jumped to 56.55%.

This jump actually supports my backwards stepwise selection process when we narrowed down our equation to one variable which was Number of Reviews.

My model equation from this output is :

Loves = 416.16 + 33.12*Number_of_Reviews + 1257.66*Rating

I really would have thought that Exclusiveness and Difference in Value Price would have been a great interaction, because if it was only exclusive to Sephora AND had a discount people would absolutely love it, but I guess not.

Interaction terms are always challenging to interpret and if I did have interaction terms that actually were significant, I would want to make sure my model was actually improving and test for stability and accuracy.

Thanks for following along!!