Abstract: | Suppose we have a linear regression model \{Y_i} = x_i^'\beta + {e_i}\](i = 1,..., n, ...), \{e_1},{e_2},...\] independent,\E({e_i}) = 0,Var({e_i}) = {\sigma ^2},(i = 1,2,...)\], and there is no such subse?quence of {ei} which converges in probability to some constant, then when the Gauss- Markov estimate \c'\beta (n)\] of a linear estimable function\c'\beta \] is not a weakly consistent estimate, there exists no weakly consistent linear estimate of \c'\beta \]. The final condition imposed on {ei} is necessary in some meaning. This greatly improved the related result in 1]. |