Features of "BAC" ("Bootstraps, Analisis en Componentes principales")
- 1. PRINCIPAL COMPONENTS ANALYSES, performed on the total covariance matrix
(total sample), or the consensus covariance matrix (subdivided sample) of:
- either (log-transformed) raw data (the program does not make the log-transformation, you have to do it if necessary).
- or size-free variables (here, the program generates size-free variables on which it performs a principal component analysis - see Darroch-Mosimann, 1985; again, this is valid only if you previously log-transformed the data).
- 2. BOOTSTRAPS for estimating confidence interval of eigenvectors.
If data are logtransformed values of distances between anatomical landmarks,
this allows to estimate the variation of the coefficients of allometry.
- 3. ANGLES between first principal components, and associated
PERMUTATION tests to estimate statistical significance.
|
