In agricultural experiment, it happens that the accidental causes such as natural disaster or pest diseases will make the field data incomplete. If this kind of data was analyzed by the ordinary procedures of the analysis of variance, then the results obtained might not be reliable. Allan and Wishart ( 1930) gave an approach to compute one missing value and constructed the concept of estimating missing data. Yates (1933) extended the approach of Allan & Wishart to several missing values by solving several equations simultaneously. Laird & Rubin presented a general approach to iterative computation of maximum- likelihood estimates when the observations can be viewed as missing data-EM algorithm. This article follows'EM algorithm'to estimate missing value(s), and to be analyzed by the ordinary analysis of variance for the potato yield of 5 different varieties in Taiwan.