The Effect of Sample Size on the Stability of Principal Components Analysis of Truss‐Based Fish Morphometrics

• DOI: 10.1577/t08-091.1
• Publication Date: 2009-04-27T20:15:35Z
• Authors: Patrick M. Kocovsky; Jean V. Adams; Charles R. Bronte
• Keywords: Morphometrics

Abstract Multivariate analysis of fish morphometric truss elements for stock identification, description new species, assessment condition, and other applications is frequently conducted on data sets that have sample sizes smaller than those recommended in the literature. Minimum size recommendations are rarely accompanied by empirical support, we know no previous minimum multivariate elements. We examined stability outcomes principal components (PCA) elements, a commonly applied method fishes, conducting PCA 1,000 resamples each 24 different ( N ; drawn without replacement) from collections yellow perch Perca flavescens (397 fish), white Morone americana (208 siscowet lake trout Salvelinus namaycush (560 fish). Eigenvalues were inflated loadings eigenvectors highly unstable first three (PCs) whenever was number P ). Stability eigenvalues increased as : ratio all but at which stable results achieved varied species. Our suggest an 3.5‐8.0 required PC2 PC3, shape. Because some our among species examined, recommend similar evaluations Results past work used where less may require re‐evaluation.