Application of multivariate statistics in a risk-based approach to regulatory compliance

The application of risk analysis as a method to ensure food safety presents significant challenges to the regulatory community, including developing sampling and regulatory scheme based on historical data that focuses attention on firms with poor compliance records. This study examines the application of multivariate statistical analysis including principle component analysis, cluster analysis, and discriminant analysis to characterize Texas feed and fertilizer firms' ability to manufacture nutritionally uniform products. Multivariate statistical results from a three year continuous data set and three variables were used to develop a sampling plan in which the best performing feed and fertilizer manufacturers were sampled at the lowest sampling percentage of the facilities. Sampling was optimized within each group to achieve the target number of total samples for the 2007 plan of work for the Office of the Texas State Chemist.