Dimension Reduction and Selection of Landmarks
A methodology of discarding landmarks that convey little shape and size information and is based on principal component analysis is considered. A sample of 100 two-dimensional “forms” described by five landmarks (three independent and two linearly dependent, with an independent perturbation component) is simulated. It is shown that the essential dimensionality of such data sets can be revealed by comparing eigenvalues of the data correlation matrix (PC values). A simple criterion to select the optimal landmark subset that uses raw data and minimizes the variance between PCs values is suggested.
KeywordsDimension Reduction Simple Criterion Size Information Landmark Position Fitted Form
Unable to display preview. Download preview PDF.
- Bookstein, F. L. 1991. Morphometric tools for landmark data: Geometry and biology. Cambridge University Press: Cambridge.Google Scholar
- Goodall, C. 1991. Procrustes methods in the statistical analysis of shape (with discussion and rejoinder). Journal of the Royal Statistical Society, Series B, 53: 285–339.Google Scholar
- Mardia, K. V., J. T. Kent and J. M. Bibby. 1979. Multivariate analysis. Academic press: London:Google Scholar
- Reyment, R. A. 1990. Reification of classical multivariate statistical analysis in morphometry. Pages 123–144 in F. J. Rohlf and F. L. Bookstein, (eds.), Proceeding of the Michigan morphometric workshop. University of Michigan Museum of Zoology Special Publication 2.Google Scholar
- Reyment, R. A. 1991. Multidimensional palaeobiology. Pergamon Press: Oxford.Google Scholar