PCA analysis

In analysis of galaxy-observation data, the term PCA analysis is used for a type of method of determining their morphology (galaxy classification), which can be suitable for automatic classification. It uses a type of principal component analysis (PCA), which is a general statistical method used to devise independent random variables that explain the variation. Specifically, it uses a type that handles weighted data (weighted principal component analysis, WPCA, using weighting to compensate for biases) and is designed to handle non-linear relationships (non-linearly principle component analysis, NLPCA, which applies PCA to data transformed such that the desired relationship is linear).

For the latter, the term principal curve (P-curve) refers to the equivalent of a principle component, but is a smooth curve that captures the desired qualities. A transformation must be devised for the purposes of the analysis, though some common techniques are established. PCA is often used in these techniques.

For galaxy classification, parameters associated with the observation of many galaxies are taken as a data set, and a principal curve is derived such that the transform to place a galaxy on the principal curve reveals its morphology.