In the weighted analysis of cell means, weighting was determined by the nij . This resulted from the assumption of constant variance for the ϵ ij k; that is, Var(ϵ) = Iσ 2 . Suppose the variances for...

In the weighted analysis of cell means, weighting was determined by the nij . This resulted from the assumption of constant variance for the ϵ_ijk; that is, Var(ϵ) = Iσ². Suppose the variances for the observations as well as the numbers of observations differed from cell to cell. Let the variance of cell (i, j) be σ²
_ij. What would be an appropriate weighting for the weighted analysis of cell means? How would you determine numerical values for the weights?