Overview of selected multivariate statistical methods and their use in phytopathological research

ABSTRACT To disentangle the nature of a pathosystem or a component of the system such as disease epidemics for descriptive or predictive purposes, mensuration is conducted on several variables of the physical and chemical environment, pathogenic populations, and host plants. For instance, it may be desired to (i) distinguish pathogenic variation among several isolates of a pathogen based on disease severity; (ii) identify the most important variables that characterize the structure of an epidemic; and (iii) assess the potential of developing regional scale versus site-specific postmanagement schemes using weather and site variation. In all these cases, a simultaneous handling of several variables is required, and entails the use of multivariate statistics such as discriminant analysis, multivariate analysis of variance, correspondence analysis, and canonical correlation analysis. These tools have been used to varying degree in the phytopathological literature. A succinct overview of these tools is presented with cited examples.