^{page 1224 note 1} These comments refer principally to the first part of Achen's paper, to p. 1226; most of them apply as well to his multiple regression analysis, pp. 1227ff., but the latter does assume heterogeneity. This is purchased at the considerable price of assuming each random variable is realized only once (all sample sizes are one), introducing—obviously—a really desperate problem of sampling variation and estimation. It seems to me almost incredible that anyone would use estimates derived under such assumptions as the dependent variable in a multiple regression, but the whole issue is mooted by the fundamental problems in Achen's model that are discussed below. The sampling problem itself is considered in some detail in the Hunter and Coggin letter.