Multivariate Repeated Measures Analysis of Sugarcane Breeding Data.

The economics of sugarcane production depends to a large extent on the yields of the cultivars planted and the ability of cultivars to produce economic yields over several crop-years.  Sugarcane breeding combines traits that impart high sugar yield and ratooning ability into a single genotype.  During breeding, yield, quality and agronomic variables are measured from the same plot over several crop-years, generating multivariate repeated measures data. Sugarcane breeders generally use univariate analysis, which assumes independence of the variables.  Multivariate repeated measures analysis account for the correlation among variables as well as across crop-years.  The objectives of this study were to determine the multivariate effects and covariance structures for repeated measures, and to compare genotypes using multivariate profile analysis.  Data on yield traits (cane and stalk dry matter), quality traits (sucrose % and fiber %), and agronomic traits (height and diameter) were collected from 16 genotypes that were planted in five blocks and harvested over eight crop-years at the Mkwasine and Triangle locations in the South East Lowveld of Zimbabwe.  A multivariate mixed model with a UN@CS covariance structure was the most appropriate because of simplicity and better control of Type II error.  Genotype 10 (highest sugar yield) and genotype 16 (the standard) were compared using multivariate profile analysis.  The yield and quality profiles for genotypes 10 and 16 were significantly not parallel, not coincident, and not level, indicating the presence of genotype by crop-year interactions, genotype differences, and fluctuations across crop-years, respectively.  Agronomic traits produced parallel profiles indicating the absence of a genotype by crop-year interaction.  Contrasts by crop-year showed that genotype 16 produced significantly (P = 0.05) greater yields than genotype 10 from the fourth to seventh crop-year.  Genotype 10 produced consistently higher sucrose %, less fiber %, and thicker stalks than genotype 16 across crop-years.