Use of ridge regression to predict yield reduction by Fusarium sp. in selected soybean cultivars

Abstract

In field studies at Georgetown and Newark, Delaware, significant differences existed among selected soybean cultivars in response to infection by Fusarium oxysporum with regard to yield, yield reduction, and frequency of recovery of F. oxysporum from roots and Fusarium spp. from the rhizosphere. Yield reduction was predicted with both an ordinary least squares model selected with a stepwise procedure and with a ridge regression model. The ridge model was more stable, used fewer predictors and was more consistent in predicting yield depression than the model selected with a stepwise procedure. Various statistics were compared to substantiate the superiority of the ridge model with plant pathological data.