The fact that measured lice were not returned to their hosts is an interference in the system that may reduce or eliminate temporal correlations that would otherwise be expected. One mathematically viable covariance model that is in line with virtual randomization, replaces each occurrence of \(t \wedge t^{\prime}\) in (5.6) with the rank-one Boolean product matrix \((t>0)\left(t^{\prime}>0ight)\), so that the only non-zero temporal correlations are those associated with lineage and aviary as strictly post-baseline block factors. Fit this modified block-factor model to the PC1 response with (5.5) for the mean subspace. Which model fits better? Is the log likelihood difference small or large? An informal comparison suffices at this point.