About |
(87 K) |
Non-Replicated Testing in the Real World
|
ABSTRACT The objective of this study was to determine the utility of non-replicated evaluations in predicting line performance. Two sets of progeny rows were evaluated in non-replicated trials at the Stoneville Pedigreed Seed Company research station near Lubbock TX in 1999. A subset of these entries was evaluated in replicated trails at the same location in 2000. Correlation coefficients of yield in 2000 trails vs. visual appeal rating, unadjusted yield, and yield expressed as a percentage of the nearest check in 1999 were generally not significantly different from zero. Correlation coefficients of yield in 2000 vs. adjusted yield calculated from Modified Augmented Design analysis in 1999 resulted in r-values of 0.20 (n=84, p=0.08) and 0.24 (n=132, p<0.01) for progeny row Set 1 and Set 2, respectively. R-values for storm resistance ratings and for fiber length, strength and micronaire in 1999 vs. 2000 were >0.50 (p<0.01) for both progeny row sets. R-values for maturity rating in 1999 vs. 2000 were 0.37 (n=84, p<0.01) and 0.17 (n=132, p=0.06) for Set 1 and Set 2, respectively. In this experiment, visual appeal ratings were worthless for predicting future yield performance. Measured yield was better at predicting future yield, but r-values were still rather low, given the resources required to collect the data. Fiber quality, storm resistance, and to a lesser extent, maturity evaluated in non- replicated trials were fairly accurate predictors of subsequent performance for these traits. |
©National Cotton Council, Memphis TN |
Document last modified May 20, 2002
|