Published online by Cambridge University Press: 27 March 2009
A missing value technique for the analysis of data where repeated measurements are taken on each experimental unit is proposed. The method suggested, which is a natural multivariate analogue of the well-known technique for univariate analyses, fits those missing values which minimize the determinant of the sums of squares and products matrix and provides the necessary downward adjustment to the residual degrees of freedom. A simulation exercise provides evidence that the method is fairly reliable.