The MVPMODEL Procedure

Using the MVP Procedures

There are two primary scenarios for using the MVP procedures:

  1. To determine whether a process is stable, you can construct $T^2$ and SPE charts from an existing set of process measurements (this is referred to as a Phase I analysis). First, build a principal component model with the MVPMODEL procedure, saving the measurements and the computed observationwise statistics (including $T^2$ and SPE) in an OUT= data set. Then specify this data set as a HISTORY= input data set for the MVPMONITOR procedure to create $T^2$ and SPE charts. Contribution plots indicate which of the original variables are involved in unusual variation displayed by the $T^2$ and SPE charts. Follow-up action might be needed to adjust the process and eliminate unusual variation signaled by the charts.

  2. To detect changes in a stable process, you can construct $T^2$ and SPE charts from newly acquired data by using the principal component model developed from previous data (this is referred to as a Phase II analysis). You can save information about the model in the OUTLOADINGS= data set created by the MVPMODEL procedure. Specify this data set as a LOADINGS= input data set and specify the new data as a DATA= input data set to create $T^2$ and SPE charts with the MVPMONITOR procedure.