2. Suppose that Yi ~ Poisson(Mi) and you wish to regress Y on a set of predictors X using a log-linear model log Mi = Xi / However, you only observe a censored version of Yi, Zi = I[Y; > 0], that is, you only know if Yi is zero (Zi = 0) or not (Zi = 1). Show that you can still estimate S by regressing Z on X using the complementary log-log link; that is, with log (- log(1 - E[Zi])) = xi Y the coefficients S and y coincide