Application of Principal Components Analysis to the Study of Silicate Crystal Structures.

摘要

Principal components analysis (PCA) is a quasi-statistical technique for transforming a given set of variables or observations into a new set of linearly transformed variables. The new set has the advantage that most of the variance of the original observations can be accounted for with fewer variables, thus simplifying interpretation of the variance. PCA has been applied to (1) micas using only interatomic distances and (2) clinopyroxenes using both interatomic distances and angles. In all cases, at most three new variables were necessary to account for 90 percent or more of the original total variance. Each of these new variables can be interpreted in terms of distortions of the crystal structures of the two groups of minerals. One of the advantages of PCA is the quantitative evaluation of specific crystal structure distortions.