Data containing many variables are often collected in weed science research, but until recently few weed scientists have used multivariate statistical methods to examine such data. Multivariate analysis can be used for both descriptive and predictive modeling. This paper provides an intuitive geometric introduction to the more commonly used and relevant multivariate methods in weed science research, including ordination, discriminant analysis, and canonical analysis. These methods are illustrated using a simple artificial data set consisting of abundance measures of six weed species and two soil variables over 12 sample plots.
