Improvement Insights Blog
To P or Not to P
Most Six Sigma statistics rely on p values. But statisticians have been raising alarm about the problems with p values. Here’s what’s going on:
“Hi, I’m Jay Arthur, author of “Lean Six Sigma For Hospitals” and the QI Macros [software].
“I get these emails about different courses that are being offered (I think one on Coursera) but it was one on statistics and it was “To P or Not to P That is the Question.”
“Have you been watching anything in the statistics world? Statisticians think that P values can be misleading and potentially dangerous. You need to look at more things than just your p value. Is it above 0.05 or below 0.05? There are times when 0.20 would be really good in terms of things that you’re trying to do, like get more traffic to your website or something.
“This is one of those things that’s pretty funny, but there’s been a lot of argument, I think, in the statistical community that p values are dangerous and you need to factor in other thoughts, not just salute the p value and go on your merry way. I’ll let you go research that online, but p values are good and… can be dangerous.
“To p or not to p, that is the question. That’s my Improvement Insight for this week. Let’s go out and improve something this week and make sure the p value is telling us what we need to know.”
