Hybrid principal component analysis in multivariate allometric regression

In biological data from allometry studies, the largest eigenvalue is typically dominant, and the gaps between minor eigenvalues are often narrow.

Such proximity among small minor eigenvalues can lead to instability in statistics based on their corresponding eigenvectors.

This study derives the asymptotic normality of the hybrid principal component analysis estimator of the leading principal eigenvector in the multivariate allometric regression model and proposes a test based on a geometric statistic for the parallelism between the regression direction and the principal component direction that avoids this instability.

Using the hybrid principal component analysis framework, we analyze the well-known painted turtle carapace data and confirm previously reported results on the allometric extension relationship between female and male turtles.