4.5. Suppose data are an incomplete random sample on Y1 and Y2, where Y1 given y ¼ ðm1, s11, b2013, b2113, b2313, s2213) is Nðm1, s11) and Y2 given Y1 and y is Nðb2013 þ b2113Y1 þ b23:13Y2 1 , s2213). The data are MCAR, the first r cases are complete, the next r1 cases record Y1 only, and the last r2 cases record Y2 only. Consider the properties of Buck's method, applied to

(a) Y, and Y2, and

(b) Y, Y, and Y = (so that Y, has the same pattern as Y, and is imputed from the regression of Y, on Y, Y2), for deriving estimates of (i) the unconditional means E(Y10) and E(Y210), and (ii) the conditional means E(YY, 0), E(YY, 0), and E(Y2|Y, 0).