Centering and Scaling :: SAS/QC(R) 12.3 User's Guide

Centering and Scaling

By default, the variables are centered and scaled to have mean 0 and standard deviation 1. Without centering, both the mean variable value and the variation around that mean are involved in selecting principal component loadings. Scaling serves to place all process variables on an equal footing relative to their variation in the data. For example, if Time and Temp are two of the process variables, then scaling says that a change of $\mr {std}\left(\mbox{\Variable{Time}} \right)$ in Time is roughly equivalent to a change of $\mr {std}\left( \mbox{\Variable{Temp}}\right)$ in Temp.

The formulas that are used to compute the variation in the different centering and scaling cases are defined in the section “Definitional Formulas” in Appendix A: Special SAS Data Sets in SAS/STAT 12.3 User's Guide. The definitional formula that is used when either the NOSCALE or COV option is specified is the COV formula. The definitional formula that is used when the NOCENTER option is specified is the UCORR formula. The definitional formula that is used when both the NOCENTER and NOSCALE options are specified is the UCOV formula. The default definitional formula, when no centering or scaling options are specified, is the CORR formula.

The MVPMODEL Procedure

Centering and Scaling