Link to home

Multivariate Mixed Linear Model Analysis of Longitudinal Data: An Information-Rich Statistical Technique for Analyzing Plant Disease Resistance

November 2012 , Volume 102 , Number  11
Pages  1,016 - 1,025

Yogasudha Veturi, Kristen Kump, Ellie Walsh, Oliver Ott, Jesse Poland, Judith M. Kolkman, Peter J. Balint-Kurti, James B. Holland, and Randall J. Wisser

First and ninth authors: Department of Plant and Soil Sciences, University of Delaware, Newark 19716; second, fourth, and eighth authors: Department of Crop Science, North Carolina State University, Raleigh 27695; third author: Department of Plant Pathology, Ohio State University, Wooster 44691; fifth author: Hard Winter Wheat Genetics Research Unit, U.S. Department of Agriculture–Agricultural Research Service (USDA-ARS), Manhattan, KS 66506; sixth author: Department of Plant Pathology and Plant-Microbe Biology, Cornell University, Ithaca, NY 14853; seventh author: Department of Plant Pathology, North Carolina State University, Raleigh; and seventh and eighth authors: Plant Science Research Unit, USDA-ARS, Raleigh, NC 27695.

Go to article:
Accepted for publication 12 July 2012.

The mixed linear model (MLM) is an advanced statistical technique applicable to many fields of science. The multivariate MLM can be used to model longitudinal data, such as repeated ratings of disease resistance taken across time. In this study, using an example data set from a multi-environment trial of northern leaf blight disease on 290 maize lines with diverse levels of resistance, multivariate MLM analysis was performed and its utility was examined. In the population and environments tested, genotypic effects were highly correlated across disease ratings and followed an autoregressive pattern of correlation decay. Because longitudinal data are often converted to the univariate measure of area under the disease progress curve (AUDPC), comparisons between univariate MLM analysis of AUDPC and multivariate MLM analysis of longitudinal data were made. Univariate analysis had the advantage of simplicity and reduced computational demand, whereas multivariate analysis enabled a comprehensive perspective on disease development, providing the opportunity for unique insights into disease resistance. To aid in the application of multivariate MLM analysis of longitudinal data on disease resistance, annotated program syntax for model fitting is provided for the software ASReml.

© 2012 The American Phytopathological Society