Consider the following statement taken from Sect. 5.5. Any credible analysis that accommodates the virtual randomization must be invariant with respect to the same group, and similar remarks apply to numerical conclusions regarding temporal trends, variance components or other effects. Invariance in this setting means that each distribution in the model is exchangeable, or invariant with respect to sex-preserving baseline permutations. This is a demanding standard, and it is possible that subsequent statements in that same section may not live up to it. Show that the model-formula Host: as. factor (Time), which is related to Table 5.4, corresponds to a set of vectors, some of which are not group-invariant. Investigate the implications, particularly for time zero.