Inference about covariances under missing values
Serge B. Provost · Environmetrics · 1991
Abstract The likelihood ratio criterion for testing the mutual independence ofqsubvectors of anl‐dimentional normal vector on the basis of a sample of sizeNonlcomponents and of an incomplete sample of sizeMonrcomponents (r < l). This is equivalent to testing the hypothesis that each variable in one subvector is uncorrelated with each variable in the other subvectors. This asymptotic distribution of the test statistic is given.