Hypothesis Testing: Remember There are BOTH Statistical and Practical Problem and Solution Statements

      Recently I went through a series of (free) online Lean Six Sigma videos by Matt Hansen at http://www.statstuff.com/.  The videos are excellent and in the series on hypothesis testing, I was reminded again how important it is to remember there are both statistical AND practical steps.  Many of us get so enthralled with the power of statistical testing from the ease of desktop software nowadays, we forget the practical implications of the findings.

      Matt suggests a four step process for hypothesis testing:

1.     Practical Problem: State the problem as a practical yes/no question.

2.     Statistical Problem: Convert the problem to an analytical question identifying the statistical tool/method.

3.     Statistical Solution: Interpret the results of the hypothesis test with an analytical answer.

4.     Practical Solution: Interpret the analytical answer in a practical way.

My experience has been failure sometimes occurs between steps 3 and 4 when it is time to ‘bring it on home.’  The statistical test has performed its’ duty when interpretation allows us to either reject or fail to reject the null hypothesis.  What happens is we don’t subject the statistical solution to a common sense dose of a practical solution.  Subject matter expert interpretation is essential at this step.

So don’t check your common sense at the door when you venture into the world of hypothesis testing.  This tool is one of the key members of the six sigma tool kit.  But you don’t want to perform all the pre-work to gather appropriate data, perform a correct hypothesis test and then blow it in the home stretch with false interpretation.