U.S. flag

An official website of the United States government

Dot gov

Official websites use .gov
A .gov website belongs to an official government organization in the United States.


Secure .gov websites use HTTPS
A lock ( ) or https:// means you’ve safely connected to the .gov website. Share sensitive information only on official, secure websites.


Main content area

Application of the factor analytic model to assess wheat falling number performance and stability in multienvironment trials

Stephanie M. Sjoberg, Arron H. Carter, Camille M. Steber, Kimberly A. Garland Campbell
Crop science 2021 v.61 no.1 pp. 372-382
alpha-amylase, data collection, genotype, genotype-environment interaction, linear models, risk, seed development, starch, universities, wheat
A factor analytic model was used to characterize data generated with the Hagberg–Perten falling number (FN) method, a measure of wheat (Triticum aestivum L.) quality influenced by genotype‐by‐environment interactions. The FN method detects starch degradation due to the presence of the enzyme α‐amylase in wheat grain such that a low FN indicates high α‐amylase activity and high risk of poor end‐product quality. Because farmers receive severe discounts for low FN, FN data have been collected over multiple years for the Washington State University multilocation variety trials to help farmers and breeders identify lower risk varieties. Analysis of these data to objectively rank varieties is challenging because the dataset is unbalanced and because FN is subject to complex genotype‐by‐environment interactions. Low FN can result from environmental differences at multiple stages in grain development because there are two major causes of α‐amylase accumulation in grain, late‐maturity α‐amylase (LMA) and preharvest sprouting (PHS). A five‐factor analytic model extracted explicit measures of overall performance and of stability in variable environments from historical FN data from the multilocation trial, providing a basis for breeding and planting decisions. Whereas a linear model explained 70.3% of the variation, the five‐factor analytic model accounted for 92.5% of variation in the data. Examination of factor loadings enabled us to separate environments and genotype response to either PHS or LMA, specifically. This is the first application of a factor analytic model to evaluate the end‐use quality trait FN, providing a method to rank varieties for grower decisions and breeder selections.