0

I'm trying to determine the variance of the error term without the residuals or the original X or Y vectors. I do have SSR. Is sigma squared of the error term just SSR/n-1?

For some reason I can't find this among tutorials. Thanks

DJF
  • 1
  • 2
    Please tell us what you mean by "SSR." – whuber Sep 04 '20 at 12:54
  • Sum of squared residuals. Some books call it SSE for sum squared errors. – DJF Sep 08 '20 at 02:41
  • And other books use "SSR" for the sum of squares of the *regression* (which is related to $1-R^2$). That's why it's crucial to explain your abbreviations! Regardless, the usual unbiased estimator is SSR/(n-1) only when -- *including any intercept* -- there is just one regressor. The general answer is found in many threads here: see https://stats.stackexchange.com/search?q=regression+error+variance+n-p . – whuber Sep 08 '20 at 14:25

0 Answers0