Misconceptions About Hypothesis Testing

While the p-value is considered the most popular way to reject a null hypothesis, there is

still risk in relying on this alone. The p of 0.05 is the value at which most draw the line

in accepting or rejecting the null in hypothesis testing. However, it's important to

remember that a p of 0.05 means that there is still the possibility of making a false

assertion five percent of the time. This misconception can cost the reliability of

hypothesis testing in some Six Sigma Projects.



By definition, hypothesis testing is being able to determine whether one group of data is

different from another. The null hypothesis is key to this process as it always states that

the variables under study in the hypothesis testing makes no difference. This puts the

burden on those testing the data to find another way to prove that the variables do make a

difference. If users reject the null, the variables do indeed make a difference in the

outcome measures, which is generally the goal of the Six Sigma project.

When rejecting the null hypothesis, there's more that should be considered than the p-value.

While this value is generally a reliable number, five percent is certainly not zero percent.

This means that there is a chance you're making an assumption on data that simply isn't

correct. When an incorrect assumption on the null hypothesis is made, a Type 1 error occurs

and this will cost you your project. Additionally, strictly adhering to this p-value rule

may mean you miss the "critical X" in your project even if it's contained in the data right

before your eyes. Missing this means the aim of your project won't be corrected, regardless

of the efforts put forth by the DMAIC process. Missing the critical X is known as a Type II

error. Ignoring the potential of these errors is detrimental to your project.

Avoiding Type II errors is key to correct hypothesis testing and can be safely done by

increasing your statistical power. This power is the ability to detect a true difference,

which aids you in appropriately rejecting your null hypothesis. You can increase statistical

power by having an appropriate sample size, and determining this before data collection

occurs. Designing your data collection to give you continuous data may also increase this

power. This type of data contains more degrees of freedom than attribute data. These degrees

of freedom allow you to more easily detect differences during hypothesis testing.

It's important to keep in mind that the p-value makes projection completion easier but might

cost you your Six Sigma project reliability at the same time. Increase your statistical

power to prevent this from happening.

Aveta Solutions - Six Sigma Online http://www.sixsigmaonline.org offers online six sigma

training and certification classes for lean six sigma, black belts, green belts, and yellow

belts.

Article Source: http://EzineArticles.com/?expert=Tony_Jacowski

Article Source: http://EzineArticles.com/6179932