Facility and management characteristics of large Upper Midwest dairy herds clustered by Dairy Herd Improvement records

Abstract

Principal component analysis (PCA) is used on datasets with vast numbers of numeric variables and limited observations (e.g., dairy herd improvement [DHI] data) for unbiased selection of uncorrelated variables that describe the largest amount of variance. Cluster analysis (CA) divides objects of interest (e.g., dairy herds) into groups on the basis of similarity in multiple characteristics simultaneously. The aims of this project were to develop a novel method for discovering important DHI variables by use of PCA and then grouping herds by those variables via CA, and to survey herds to determine herd management characteristics of each group.