Can someone please help me with what the proper denominator should be for the MSE (mean squared error) for a regression? My textbook, everything I read online, and the Excel regression output indicate it should be "n-p-1" (n being number of observations and p being number of independent variables). Formula is SSE/(n-p-1). All made sense to me and I thought I was ok.

When the professor reviewed our first one variable linear regression problem online with us, she indicated that "technically" the denominator for MSE should be just "n". She re-did the calculation using just "n" and indicated to check with the professor whose class it was (she was just filling it for one session).

We were next assigned linear regression problems that had a data set of sales numbers. For example, it was 48 monthly sales for a department store. Run a regression on these numbers. The one assignment had us use dummy variables for the monthly seasonality. I was again told that "n" would be the correct denominator for the MSE (and not 35 as Excel used in the ANOVA table 48-12-1=35).

I have asked the full time professor for the class and I haven't really gotten an explanation that explains why to me. I was basically told just "n" is correct and other statistical software all use "n" (and not n-p-1).

Any help with greatly be appreciated as I find this really confusing.

Thanks