Chapter 7. Correlation Tests
What you will learn in this chapter is how to discover the key process input variables (KPIVs) that may have caused a change in a process or product. For that, we will be doing correlation tests. In some Six Sigma classes, regression analysis is used to find correlations. A mathematical curve is fit to a set of data and techniques are used to measure how well the data fit these curves. The curves are then used to test for correlations.
These methods require a high degree of skill and generally are not friendly to those who are not doing this kind of analysis almost daily. Thankfully, most Six Sigma work can be done using the tools already covered, as long as we are willing to do some visual examination of data and their related graphs. Correlation tests are used primarily in the Define, Analyze, and Improve steps of the DMAIC process.
Something changed in a process or product and we would like to discover the KPIV(s) that caused it. Time and position are the critical factors in doing the analysis.
NOTE
Correlation Tests
Manufacturing Do a time plot showing when a problem first appeared or when it comes and goes. Do similar time plots of the KPIVs to see if a change in any these variables coincides with the timing of the change. If so, do a controlled test to establish the cause-and-effect relationship for that KPIV.
Sales and Marketing For periods of unusually low sales activity, do a time plot showing when the low sales periods started and stopped. Do similar time plots of the KPIVs to see if a change in any these variables coincides with the low sales period. If so, do a controlled test to establish the cause-and-effect relationship for that KPIV.
Accounting and Software Development Do a time plot of unusual accounting or computer issues. Do similar time plots of the KPIVs to see if a change in any these variables coincides with the issues. If so, do a controlled test to establish the cause-and-effect relationship for that KPIV. The people in these areas respond well to this type of analysis.
Receivables, Insurance, etc. Identify periods when delinquent receivables are higher than normal or the frequency of claims is unusual. Then do a time plot of the problem and the related KPIVs. For any variable that shows coincident change, check for cause-and-effect relationships with controlled tests.