The variance of the error ( ) is a parameter. In other words, there’s some true, but unknown, value for this in the population as a whole. The MSE, however, is an estimate of the error variance. Like all estimators, you never know exactly what value will be derived, because it varies from sample to sample. The assumption that the error is normally distributed implies that the MSE and the estimated variances of the coefficients are the square of a normal, so they have a
chi-squared (χ2) distribution with n – p – 1 degrees of freedom.