ABSTRACT
It is worthwhile to step back and put this and the preceding section together. In presentations that argue for a stratified analysis, the first step is to exhibit a joint model in which the parameter of interest can be unambiguously identified. In our case, this would likely be pL2, although it would seem that p!2 might also be of interest. The development then shows that pi.2 is mis-specified in the marginal analysis. The culprit here is usually identified as a stochastic relationship between Xi and x2. The demonstration then turns to the stratified analysis, showing that Pi.2 is correctly specified.
