Predicting double cross yields in corn

Abstract

The yield of shelled corn in bushels per acre for 74 single crosses and 20 double crosses grown on fields of the Texas Agricultural Experiment Station near Angleton, College Station, and Tyler, Texas, over an 11 year period was analyzed using the least-squares method of fitting constants without and with covariates. Constants without covariates from the single cross data were used to predict the constants for double crosses. The predicted constant for a double cross was the mean of the 4 non-parental single crosses made from the 4 inbred lines used in the parentage of the double cross. The correlation coefficient between the predicted and actual double cross constants was .5110 with 18 degrees of freedom and was significantly different from zero. A scatter diagram for predicted constants and actual constants shows that when only 50% of the double crosses having the highest predicted yield are produced for testing some high yielding double crosses can be ignored. There was a significant number of low yielding double crosses identified as such that would never be produced for testing. The least-squares method of fitting constants without covariates merits usage for predicting double cross yields. Single and double cross yields were adjusted by 12 and 8 meteorological covariates taken from 4 phenological periods of the corn plant. The 12 and 8 covariate analyses had correlation coefficients between predicted and actual double cross yields of -.5433 with 11 degrees of freedom and -.0262 with 11 degrees of freedom respectively. Neither correlation coefficient was significantly different from zero. ...