Estimation of Missing Values for the Analysis of Incomplete Data

Missing Data in Quantitative Designs

“…Wilkinson (1958Wilkinson ( , 1970 and Tocher (1952) provide examples of methods which can require considerable computation owing to matrix manipulation and inversion. An analysis of covariance would also involve a great deal of computation, especially with more than one missing value.…”

Missing at Random ‐ What Does It Mean?

“…In some articles such as those concerned with the multivariate normal (Afifi & Elashoff, 1966; Anderson, 1957;Hartley & Hocking, 1971;Hocking & Smith, 1968;Wilks, 1932), "missing at random" seems to mean that each item in the data matrix is equally likely to be missing. In other articles such as those dealing with the analysis of variance (Hartley, 1956;Healy & westmacott, 1956;Rubin, 1972;Wilkinson, 1958), "missing at random" seems to mean that observations of the dependent variable are missing without regard to the actual values that would have been observed. Similarly, "missing at random lt apparently can mean missing according to a preplanned experimental design (Hocking & Smith, 1972-;Trawinski & Bargmann, 1964).…”

Iterative Procedures for Missing Values in Experiments

“…The values m = 0.2(0.2)3.0 and m = 1.45(0.05)1.55 were used for the following randomised blocks variate, given by Wilkinson (1958), with n' = 3; missing values are denoted by asterisks: The obtained values of X' were used to draw the curves in Figure 1, where X', on a logarithmic scale, is plotted against m. Successive curves (counting from top to bottom on the left side of the Figure) correspond to successive iterations. Thus, for example, if X' < 0.005 is taken as the stopping criterion, five iterations are needed if m = 1, four if m = n/E = 15, and three if m = 14.…”