About this lesson
There are many different statistical tests that can be used with hypotheses. The tests shown in this module are the most commonly used tests in Lean Six Sigma. The decision to choose a test is based upon many factors associated both with the data and the nature of the hypothesis statements.
When to use
Use hypothesis tests to provide statistical confidence concerning your assumptions with regards to root causes of the problem and the performance of the improved process.
The decision tree shown below should be used for selecting which hypothesis test to use. In some cases, more than one test could be used, so the most commonly used one is shown. In several of the non-normal categories, there are multiple options. For those, choose the test that best matches the shape and characteristics of the data. In some cases, a preliminary hypothesis test must be done prior to the final test because the preliminary test will provide some key insight into how to complete the final test (F Test, Bartlett’s, Levene’s).
Hints & tips
- Follow the paths of the decision tree, do not skip a decision.
- In some cases, a test may be applicable for multiple conditions (ANOVA.) However, this decision tree attempts to show the simplest test for each of the conditions.
- Many of the Normal tests are somewhat forgiving if the data is close to normal. However, if the data is definitely non-normal, use a non-normal test.
- Excel is able to conduct virtually all of the normal tests and none of the non-normal tests.
- Minitab will perform all the tests listed (and many more that are not listed.)
Lesson notes are only available for subscribers.