Tail behavior of the least squares estimator is considered with a random design and with infinitely increasing n and with gamma equal to the population extreme. We find a fairly good behavior of LSE for normal F.
