Pl. Cornelius et J. Crossa, Prediction assessment of shrinkage estimators of multiplicative models formulti-environment cultivar trials, CROP SCI, 39(4), 1999, pp. 998-1009
Multiplicative statistical models such as the additive main effects and mul
tiplicative interaction model (AMMI), the genotypes regression model (GREG)
, the sites regression model (SREG), the completely multiplicative model (C
OMM), and the shifted multiplicative model (SHMM) are useful for studying p
atterns of yield response across sites and estimating realized cultivar res
ponses in specific environments. Traditionally the series of multiplicative
terms is truncated at some point beyond which further terms are believed t
o have little statistical significance or predictive value. Shrinkage estim
ators have been advocated as a model fitting method superior to model trunc
ation. In this study, by data splitting and cross validation, rye evaluated
the predictive accuracy of (i) truncated multiplicative models, (ii) shrin
kage estimators of multiplicative models, (iii) Best Linear Unbiased Predic
tors (BLUP) of the cell means based on a two-way random effects model with
interaction, and (iv) empirical cell means in one wheat [durum (Triticum tu
rgidum L. var. durum) and bread (Triticum aestivum L.)] and four maize (Zea
mays L.) cultivar trials, with and without adjustment for replicate differ
ences within environments. Shrinkage estimates of multiplicative models wer
e at least as good as the better choice of truncated models fitted by least
squares or BLUPs. Shrinkage estimation yields potentially better estimates
of cultivar performance than do truncated multiplicative models and elimin
ates the need for cross validation or tests of hypotheses as criteria for d
etermining the number of multiplicative terms to be retained. If random cro
ss validation is used to choose a truncated model, data should be adjusted
for replicate differences within environments.