How does one obtain the best estimate of the variance of observations around the fitted model in SPSS?

My thinking is that the best estimate of the variance of observations is the F test statistic value generated by SPSS because it is the ratio of the two sample variances found by dividing between group variances by within group variances.

Am I correct in my thinking? Or is the best estimate of the variance the residual mean square value?