Main content area

Predicting loaf volume for winter wheat by linear regression models based on protein concentration and sedimentation value using samples from VCU trials and mills

Laidig, Friedrich, Piepho, Hans-Peter, Hüsken, Alexandra, Begemann, Jens, Rentel, Dirk, Drobek, Thomas, Meyer, Uwe
Journal of cereal science 2018 v.84 pp. 132-141
baking, equations, loaves, prediction, regression analysis, surveys, variety trials, winter wheat, Germany
The determination of loaf volume (LV) in winter wheat by the rapid mixed test (RMT) is laborious and expensive. In Germany LV of wheat samples from a national harvest survey is predicted routinely by a linear regression equation introduced by Bolling in 1969 with predictor variables protein concentration (PC) and sedimentation value (SV). Breeding progress generated varieties with improved quality so that the predictive power of the Bolling's equations, which were updated, decreased steadily. The objective of this study is to evaluate linear regression models based on PC and SV with improved accuracy using data from mill samples and from official variety trials (VCU) including varieties with quality ratings A, E and B. As criterion for the prediction accuracy we used the root mean squared deviations sMSD of observed and predicted LV. Two basic regression models were used across quality groups and group-wise fashion. The average prediction accuracy of Bolling's equations (sMSD = 77.0) was considerably lower than the derived regression options (41.5 < sMSD < 48.5). This study has shown that LV can be predicted in a fast way with reasonable accuracy using properly updated and group-wise linear regression models taking into account time trends in regression coefficients.