MWPCA Procedure

The MWPCA procedure implements principal component analysis over sliding windows that consist of a specified number of observations. You can use this procedure to capture changes in principal components over time. Also, you can choose to perform robust principal component analysis (RPCA) on each window; in RPCA, the outliers and noise are excluded from each window before the analysis is performed.

One important application of MWPCA is for detecting relative changes in parts of a system compared to the overall system. By tracing the principal components over time, you can determine whether significant changes in the principal components can be a cause for concern.

PROC MWPCA stores the requested principal components for each window in an output table that is produced by the OUTPUT statement.