Why $\sum_{i=1}^n e_i $ is a good estimator of the measurement error variance?

32 Views Asked by At

Why the sum of the $n$ squared residuals from a simple linear regression divided by $n - 2$ might be expected to estimate the measurement error variance?