3.11. Oh and Scheuren (1983) propose an alternative to the raked estimate yrake in Section 3.3.3, where the estimated counts N*jl are found by raking the respondent sample sizes frjlg instead of fnjlg. Show that (i) unlike yrl, this estimator exists when rjl ¼ 0, njl 6¼ 0 for some j, l, and (ii) the estimator is biased unless the expectation of rjl=njl can be written as a product of row and column effects.