I have been given two formulas:

$$ \hat{\sigma}^2 = \frac{1}{n-p} (Y - X \beta)^T(Y - X\beta) $$ $$ = \frac{1}{n-p} \sum ( y_i - \hat{\beta_0} - \hat{\beta_1}x_i)^2 $$

I also have

$$ \hat{\mathrm{MSE}} = \hat{\sigma^2}\left(1 + \frac{1}{n} + \frac{(x^* - \bar{x})^2}{\sum(x_i - \bar{x})}\right) $$

where n is total number of people, p is number of paramters.

Are these formulas right for calculating MSE?

In my lecture notes, my lecturer as calculated $\hat{\sigma}^2 = 0.8$ and said thus MSE = 0.8. So does that mean I don't have to work out MSE if I have $\hat{\sigma}^2$?


The first is the mean square error of regression, the residual sum of squares divided by its degrees of freedom - which is an estimate of the population variance parameter $\sigma^2$. The second is an estimate of the mean square error of a single predicted response $x^*$. So your lecturer was calculating the MSE for regression (& equating it to $\hat{\sigma}^2$), not the MSE for a predicted response.


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.