REGULAR ARTICLE
Analysis of a complex trait with missing data on the component traits
Hans-Peter Piepho, Bettina U. Müller, Constantin Jansen
Commun. Biometry Crop Sci. (2014) 9 (1), 26-40.
ABSTRACT
Many complex agronomic traits are computed as the product of component traits.
For the complex trait to be assessed in a field plot, each of the component traits needs
to be measured in the same plot. When data on one or several component traits are missing,
the complex trait cannot be computed. If the analysis is to be performed on data for the
complex trait, plots with missing data on at least one of the component traits are discarded,
even though data may be available on some of the component traits. This paper considers a
multivariate mixed model approach that allows making use of all available data. The key idea
is to employ a logarithmic transformation of the data in order to convert a product into a
sum of the component traits. The approach is illustrated using a series of sunflower breeding trials.
It is demonstrated that the multivariate approach allows making use of all available information in
the case of missing data, including plots that may have data only on one of the component traits
Key Words: logarithmic transformation; log-normal distribution; multiplicative model; multi-trait analysis; multivariate mixed model; yield component analysis; yield components.
