About Me

My photo
I am the founder of University Training Partners, a company that designs and delivers online Lean Six Sigma and statistics training. In this blog, I hope to share my thoughts on statistics, quality assurance, and training, drawing from my 15 years of work experience in the field, eight years as a tenured professor teaching graduate statistics classes and ten years as a full-time training developer.

Wednesday, January 27, 2016

Statistical Rules of Thumb and the 3.4 dpmo Dilemma

There are a lot of quality consultants out there touting "quick and dirty" statistical methods: easy ways to get an answer without all that inconvenient math. Short course students leave their Become a Statistician in Two Days classes armed with rules of thumb, quick reference guides, and flow charts of simplified methods. Do these methods work? Yes and no. They do perform as advertised: giving an answer without the rigor. But is the answer the right answer?

My question to you is this: In this age of Six Sigma Quality, where only 3.4 dpmo is acceptable (and remember, that is AFTER the 1.5 sigma shift in mean), how is it that we accept statistical rules of thumb to help us make process decisions?

We claim to wholly embrace this 6 Sigma standard, and then turn around and use critical values of +/-2 for our hypothesis tests, or blindly assume normality and happily proceed with our regression modeling. We measure with a micrometer and then cut with an ax.

It is assumed that statistics -- real statistics-- is too difficult a subject for us quality folks to master. Hence a "statistics without the pain" training industry has sprung up to give us "just the facts, nothing more." Why are we selling ourselves short? Sure, statistics requires math and lots of practice. So what? Aren't our customers and their satisfaction worth a little extra training and number crunching?

Join our mailing list to receive exclusive course discount codes and VIP access to webinars!



© 2016 Mary McShane-Vaughn

3 comments:

dbell2 said...

Good comments. I think our classes would be better if they were a little more math intensive (but not too much, they're still a little hard sometimes).

However, a 2-day quickie stats class would be good for us. We may not know all the rules of thumb that are often helpful. But maybe a 4-day class would be better. You could cover the rules of thumb, AND, the when and why they sometimes don't work.

HugeElvis said...

I like your statement, "measure with a micrometer and cut with an axe".
The "Stats in a Minute" course is probably good enough for production personnel, but quality professionals need the full blown version to explain the what and why to the production personnel with the short couse statistics.

leeda said...

I, too, like "measure with a micrometer and cut with an axe". That seems to happen way too often in many things.

I also agree that the short course would suffice for some, but if a professional were to rely soley on simple tools and rules of thumb, I think it would be a potentially disasterous situation and could lead to major embarassment for all.