As an example, the chart on the right illustrates the use of a Principal Components Analysis (PCA) multivariate modeling of a real process with 25 input variables.

A control chart (right) was made of all 25 variables simultaneously, with each point plotted being a linear combination of values for the 25 variables. The chart clearly shows the process going out-of-control on the 41st run.

The Contribution Score Plot (right) pin-points the variables associated with the out-of-control condition on point 41 so that action can be taken on these inputs. Now models can be generated for predictive purpose.

In addition, PLS (Projection to Latent Structures) was used to construct models relating four critical process parameters to the 25 process inputs.

In a matter of minutes, this process data was used to find critical inputs driving the four process outputs. This analysis resolved an issue with construction material that was not able to be resolved before applying these modeling techniques.