Range of Application of Inverse Variance Weighting for Using Historical Information in Stepwise Yield Screening of Plant Varieties

Abstract

Using yield data collected in the preceding steps of selection (historical information) for the evaluation of the yielding ability of test varieties is essential to increase the efficiency of stepwise yield screening of plant varieties. Selection in each step should be performed based on the average score over the preceding and current steps. Three methods of weighting can be used to calculate the average score, i.e., inverse variance (IV), field area (FA) and equal (EQ) weighting. In IV, yield data collected in the preceding and current steps are weighted by the inverse of the error variance of the corresponding steps, while those in FA are weighted by the field area (or plot numbers) allocated per variety. A simple average score is used in EQ. Of these three methods, although IV is the best theoretically, it has not been recommended previously nor is commonly used in actual screening trials because two necessary statistics, i.e., variances due to the measurement error (V) and variety × year interaction (I), may not be estimated correctly. Our studies demonstrated that the superiority of IV to the other methods persists under possible degrees of inaccuracy in the estimates of V and I. Calculation of the inverse-variance-weighted average does not require significant extra resources or expensive facilities. We therefore suggest that method IV be used with the necessary statistics V and I being estimated from the yield data obtained in the current and earlier steps. Stepwise screening thus performed is superior to non-stepwise screening where the yield of all varieties is measured in the same years and evaluated based on the average over the years.