I am attempting to model monthly retail electricity sales. To account for both the effects of seasonality and weather, I created an interaction term by multiplying 12 monthly dummy variables by the corresponding month's max temperature, such that:

Jan_dum_temp = Jan_dum * Jan_max_temp, 
Feb_dum_temp = Feb_dum * Feb_max_temp, 
Dec_dum_temp = Dec_dum * Dec_max_temp

The resulting model will be:

y(hat) = b1 + b2Jan_dum_temp + b3Feb_dum_temp + ... + b13Dec_dum_temp + 
  other explanatory variables

When adding these interaction terms to the regression, should I leave one interaction term out to avoid the dummy variable trap? If possible, please provide a source for your answer.

  • $\begingroup$ What do you think? Does your regression allow a constant? If so, what might it mean? $\endgroup$ – Henry May 23 at 14:38
  • $\begingroup$ Yes, there is a constant within my model. I do not think I should exclude one of the 12 monthly interaction terms because I have not run into the problem of perfect multicollinearity, especially after reviewing the table of 'Correlations of Parameter Estimates'. $\endgroup$ – Darlene May 23 at 14:54
  • 1
    $\begingroup$ Yes, leave it out (or else your solver will semi-randomly omit one of the months). Think about the interprettaion of Feb_dum_temp: it is a difference in effect of temperature from February to January. $\endgroup$ – AdamO May 23 at 14:58
  • $\begingroup$ My solver has not omitted one of the interaction terms when running the regression. However, when I manually omit the Jan_dum_temp interaction term and only include the following 11 monthly interaction terms, is the constant interpreted as the effect of January's interaction term? $\endgroup$ – Darlene May 23 at 15:06
  • $\begingroup$ Or maybe better asked, when omitting January's interaction term, how do I interpret the effects of January's seasonality and weather? $\endgroup$ – Darlene May 23 at 15:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.