Since the P-value of 0.075 is greater than the alpha of 0.05 … wait, stop, what am I saying? Nobody I speak to on a day to day basis knows what the hell I’m talking about. If you do, can you explain what a P-Value is?

Let me attempt to explain what it is in plain language.

# What P-Values Mean

When a statistical analysis software package runs a statistical test, it’s going to generate a ‘test statistic’ which is supposed to tell us something.

There it is, I can see the ’t’ statistic is 19.380 or the ‘chi’ statistic is 27.410.

One would have to have the brain of Einstein to be able to make sense of these test numbers and draw some understandable and accurate conclusion.

Thankfully every single one of these ‘test statistics’ comes with a P-Value, and the interpretation of the P-Value is always the same.

Now before I go on and you improvement professionals or statisticians out there jump down my throat and tell me I am not technically correct; this was not written for you. It was written specifically for those people new to business improvement who opened a book or statistical software package and wondered what that ‘P-Value’ thingy is.

P-Values sit on a continuum somewhere between 0.00 and 1.00 and they are nothing more than a measure of the ‘significance’ of something.

If the P-Value is above some cut off point, it means something we are testing is statistically insignificant. If it is below that point, we now say that something we are testing is statistically significant.

Every time we run a test we do so because we hypothesise about some idea.

For example .. I might hypothesise that gender (which is a categorical variable) has an effect on how tall a person is (which is a numerical variable). I base this on my thoughts that women are generally shorter than men.

I might then run a test to evaluate this idea and see whether we can (a) reject the idea or (b) validate that it is correct.

In this height example, we would collect a sample of data knowing that a sample might be slightly different from the overall population. We would then statistically compare the sample of heights of women against the sample of heights of men to answer our question.

Basically, the P-Value that would be generated tells us whether or not any difference we see in the heights is ‘significant’ or not. A little bit of difference between ‘samples’ would result in the generation of a test P-Value at one end of the continuum (i.e. more towards 1.00) and tell us the difference we are seeing based on a sample of data is ‘not significant’ enough to confirm our suspicions.

A lot of difference between ‘samples’ would result in the generation of a P-Value at the other end of the continuum (i.e. more towards 0.00) and tell us that the difference we are seeing is ‘significant’ and we can confirm our suspicions.

We would then base our ‘statistical conclusion’ on what the P-Value is indicating to us.

That’s it. How’d I do?

# More Information

This article was written by George Lee Sye, author of PROCESS MASTERY WITH LEAN SIX SIGMA – the best lean six sigma text book in the world today.