It is a new month, so I will pull out another article I have on hand. This one was originally written for the local ASQ newsletter. I ran it over about 4 newsletters. I also used it as a point paper at the Department of Energy. It is an overview of the use of control charts. I hope you find it interesting.
By the way, since I am pulling out old articles, and they were written "on the job" (with your taxpayer dollars), I should recuse myself from the $150 award. Hopefully, this article is a good seed for follow on articles.
Steve Prevette
Statistician for Fluor Hanford
Teacher at City University
Troublemaker at ASQ
Having fun on the Cove!
__________________
Steve Prevette
"A Passionate Statistician", ASQ CQE, Fluor Government Group
The opinion stated above does not necessarily reflect that of my employer.
I also glow in the dark when you turn out the lights.
__________________
Steve Prevette
"A Passionate Statistician", ASQ CQE, Fluor Government Group
The opinion stated above does not necessarily reflect that of my employer.
I’m enjoying your articles a lot. Here are some questions about “Life Cycle of a Trend”.
We are just about to generate a “pile O metrics” as required by TS16949.
People are having a field day with Excel to produce almost every imaginable type of graph. My personal favorite uses a 12 month moving average to hide a trend down. Some folks are very clever.
Every data point is either claimed by the owner as improvement if favorable or generates cries for immediate action by management if unfavorable.
People are absolutely dumbstruck when things suddenly get worse on the next data point after an improving “trend”.
Being a good little Quality Engineer I take the same data and whip up a control chart.
Sure enough most of the time the systems are in state of statistical control, most however operating at an unacceptable level.
A tremendous amount of passion and anger is spent explaining each point, and absolutely no effort is invested in improving the process to make an permanent change.
I am slowly trying to bring people over to using SPC charts, but it is a tough sell.
Now to the questions
I see you only show the average chart and not the range chart. Is this just to simplify for the paper?
I was always taught to look first at the range chart for increased variation, then inspect to average chart. I understood they made up a required set, both were needed.
These two charts cause much confusion to non Quality people.
Can your MW rule, trend rules and the average chart alone work the same as an average and range set?
If so, this “SPC Lite” approach is something I think I can sell to people.
What is lost by not using the range chart? Will the trend rules catch variation changes in any event?
Thanks for the articles, you are a clear thinker and an excellent explainer.
By the way, since I am pulling out old articles, and they were written "on the job" (with your taxpayer dollars), I should recuse myself from the $150 award. Hopefully, this article is a good seed for follow on articles.
If you have the most votes, your reward will then be given to the next person in line in the poll. I appreciate your contribution, but I also want folks to know that someone will get the check I send for the best article for the month.
For September submissions the amount will be US$200. If you're out of the US, as the last person was (Malta), it has to be a Western Union funds transfer. Technically I pay more - there is the transfer fee charged by Western Union - and the receiver gets a bit less - because the local agent charges for the currency conversion. Oh, well. So goes 'International Finance'!
I see you only show the average chart and not the range chart. Is this just to simplify for the paper?
I was always taught to look first at the range chart for increased variation, then inspect to average chart. I understood they made up a required set, both were needed.
These two charts cause much confusion to non Quality people.
Can your MW rule, trend rules and the average chart alone work the same as an average and range set?
If so, this “SPC Lite” approach is something I think I can sell to people.
What is lost by not using the range chart? Will the trend rules catch variation changes in any event?
Thanks for the articles, you are a clear thinker and an excellent explainer.
Cheers
When I started making control charts at Hanford, I started with xbar-R charts, but never could get over the hurdle of management. The first charts were on cycle times on maintenance work, taken 5 work packages at a time in sequence completed. But management could only talk monthly charts, and were confused by why one month had 4 data points and the next month 6. Certainly the whole concept of range was way beyond them. I compromised upon straight x-charts, or data accumulated on a time interval rather than number of data points. Statistically it is weaker, but at least it got the data in front of management. If I had my "druthers" I would much prefer to work with xbar-R, but I don't think the losses in practice are that great. Empirically, the trend rules and the MW rule appear to work well, so I think the SPC-lite-ness of this approach is not too awful.
__________________
Steve Prevette
"A Passionate Statistician", ASQ CQE, Fluor Government Group
The opinion stated above does not necessarily reflect that of my employer.
I especially like Figure 4 on page 16. I see this all the time in meetings when people react/overreact to the next months data point without considering common cause variation.
How did you explain the presence of control limits to non Quality people? I can see blank stares in the boardroom if I try it.
I think there is also a powerful bias in play that makes use of SPC difficult with business data. Leaders are under pressure to punish every downturn and claim every improvement as due to their actions. They do not want to hear that in all reality, their actions had no effect.
Have you had any success stories using these charts to push people to radically change the process to get lasting improvement?
I especially like Figure 4 on page 16. I see this all the time in meetings when people react/overreact to the next months data point without considering common cause variation.
How did you explain the presence of control limits to non Quality people? I can see blank stares in the boardroom if I try it.
I think there is also a powerful bias in play that makes use of SPC difficult with business data. Leaders are under pressure to punish every downturn and claim every improvement as due to their actions. They do not want to hear that in all reality, their actions had no effect.
Have you had any success stories using these charts to push people to radically change the process to get lasting improvement?
Caster
Davis's paper is a GREAT paper. I have met Davis and we have stayed in touch now and then. Davis does base his SPC on Medians and Median Ranges, so this is a little different twist than I use. The median loses some statistical power, but is very robust against outliers (thus good for novices).
I like to use the smoke detector analogy when it comes to control charts. You also can discuss type I and II error that way. People understand they don't want their smoke detector alarming every two minutes. I have attached a little one page cheat sheet I use with management.
For success stories, please see the presentation I have submitted to ASQ for the next quality conference. Since this paper was made on government time (as a contractor), the US Government has posted it out on the OSTI fileserver, so it does not appear to be a violation of copyright if you go there. I just can't give it to you directly.
__________________
Steve Prevette
"A Passionate Statistician", ASQ CQE, Fluor Government Group
The opinion stated above does not necessarily reflect that of my employer.
Life Cycle of a Trend - An SPC Overview
Linear Mode
