Use of regression and discriminant analyses to develop a quality classification system for hard red winter wheat

An attempt to create a segregation system that uses rapid quality detection instrumentation and represents value to millers and bakers led to the deve lopment of a single value called "dough factor." The dough factor value rep resents the amount of flour-water dough that can be produced by a given uni t of wheat. Samples of hard red winter wheat (approximate to 100/ location) collected from five Kansas country elevators during the 1995 and 1996 harv ests were evaluated for dough factor. Single kernel properties, sample prot ein content, and test weight measurements were subjected to regression and discriminant analyses for the purpose of developing a dough factor classifi cation system. Regression analysis identified kernel weight, kernel weight standard deviation, and protein as important characteristics for predicting dough factor, however, the resulting model possessed poor predictive abili ty (adjusted R-2 = 0.39). Classifying wheat into dough factor groups of <10 7, 107-112.9, and greater than or equal to 113 using discriminant analysis resulted in an accuracy of 56%, while discriminant analysis correctly place d wheat into two dough factor groups (<113 and greater than or equal to 113 ) with an 80% accuracy. Creation of a dough factor classification system us ing single kernel measures, kernel protein, and wheat cultivar correctly cl assified 86.3 and 68.8% of the wheat samples into dough factor groups <113 and greater than or equal to 113, respectively. In the dough factor group g reater than or equal to 113, cost savings associated with higher flour yiel ds and water absorption were $0.15/cwt of flour and $0.65/1,000 Ib of dough , respectively. Increases in processing efficiency for both the miller and the baker would be expected to further differentiate the value between the two dough factor groups.