Standard errors for biplots
by
Maaike Bendall
Crop and Food Research Institute, Palmerston North

When several correlated variables are measured, a biplot of the principal components is useful for tying together the various findings of the study. Discussing each variable in turn can be tedious, so a biplot can end up being the focus of a scientific paper. However, biplots are often presented without any standard errors for the loadings of the principal components. This makes it difficult to know whether the components are meaningful or the result of random variability. There is no shortage of methods for obtaining standard errors, but the relevant results for the parametric approach using the multivariate normal distribution are asymptotic and need a large sample size. Similarly, it would seem that a large sample size is desirable for the re-sampling methods. Here, different methods are compared on medium sized datasets from scientific experiments.

Date received: August 24, 2001

Copyright © 2001 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cahg-24.