# Applying the Out of Control Tests

SPC for Excel Software

SPC Training

SPC Consulting

Ordering Information

Thanks so much for reading our publication. We hope you find it informative and useful. Happy charting and may the data always support your position.

Sincerely,

Dr. Bill McNeese
BPI Consulting, LLC

#### Connect with Us

• Martin GibsonJanuary 31, 2017

Hi Bill, as always a good read.  Can you supply the references for the different e=tests please?Thank you and Happy New Year,Martin

• billJanuary 31, 2017

Thank you. Here are two references for the out of control tests:
Western Electric Company (1956), Statistical Quality Control Handbook
Douglas Mongtomery. Introduction to Statistical Quality Control

• MelissaFebruary 4, 2017

Thank you for this great read sir. i've searched for so long to find such material to support my spc procedures.  Finally a good find.  Hope you continue these. Really appreciate it.

• DaleWMarch 2, 2017

Bill, both of your references actually state that 8 successive points in Zone C or beyond (on the same side) is a signal for a non-random pattern.Using the 7 in a row rule gives us a 1 in 64 asymptotic risk of false positives if we have random data — doesn't that seem rather high to you?  Wikipedia even claims that the WECO rule is 9 in a row, which doesn't match their 1956 handbook, athough 9 in a row is more typically in use today (MINITAB default).

• billMarch 3, 2017

Hi Dale.  8 is the most common number.  The WECO rule is 8.   My book has it as 8.  But I shifted it to 7 simply so it matched the rule of seven tests: 7 trending up or down in a row, 7 above or below the average in a row.  I taught this as a good initial set of rules for operators back when we had to do things manually.  Start with that and move to the zones later and wanted the est for zone C to match the 7 in a row.
The probably of getting this pattern with 7 is .007813 – still pretty small.  With 8 it is .00396 which is closer .0027 of points beyond the control limits.  The software allows you to set the value for the number of points.

• DaleWMarch 6, 2017

Hi, Bill. Actually, the odds of seeing 7 random points in row above or below the established mean is 1 in 64.In Advanced Topics in Statistical Process Control, Donald J. Wheeler claims to personally prefer “nine in a row.” He mentions that Irving Burr expresses the decision rule as “8 to 10 in a row (take your choice).” Wheeler notes “It should be added that virtually all modern authors agree that Grant’s rule [seven in a row] is not sufficiently conservative for good practice.”It’s good to have flexible software.

• billMarch 6, 2017

Thanks for the reply.  7 in a row above the average is .5^7 which is 1 in 128.  Unless I am missing something there.

• DaleWMarch 7, 2017

Don't forget the either part of the rule.  1 in 128 for 7 above <em><strong>plus</strong></em> 1 in 128 for 7 below equals 1 in 64 for 7 in a row on either side of the centerline.

• billMarch 7, 2017

I will respectively disagree on that.  You are looking at 7 in a row above the average or 7 in a row below the average.  Not the sum of both.

• DaleWMarch 7, 2017

“If two events are mutually exclusive, then the probability of either occurring is the sum of the probabilities of each occurring.”As an experiment, I ran the 7-in-a-row-on-same-side-of-center-line rule only on a million random normal numbers using Minitab. It took a couple of minutes, but it flagged 15,617 points, which is really close to the theoretical 1 in 64 value of 15,625.

• eveSeptember 2, 2019

Hi!Could you explain further why only the "beyond limits" rule applies to the Moving Average?Planning to create a control chart implementing WECO rules on a moving average.Thanks!

• billSeptember 2, 2019

It is because you are reusing data with the moving averages. For example, each result is used in a subgroup mutliple times.  This means that the data are correlated – they are being reused. So the runs don't apply.  Only point beyond the control limit.   Moving average/range charts often look like cycles.

• NickJuly 4, 2022

How would you establish control charts for processes for which it is hard to assess whether they are in control or not? When talking complaint data, it is not straightforward to estimate whether it is in control or not (how many complaints are 'normal'?). I currently have a dataset which is likely out of control (many complaints), but I'm not sure how to establish a good mean and standard deviation to prove this using a control chart, as the data is so variable. If I take months with many complaints as part of the data, these months will hugely influence the mean and SD, and thus 'allow' for more complaints in subsequent months, which is where I'm getting quite confused. Thanks!

• billJuly 4, 2022

Can you email the data and I will take a look at.  Hard to know without seeing the data.  Thanks.  [email protected]