# p-Charts With Straight Control Limits

Six Sigma – iSixSigma Forums General Forums Tools & Templates p-Charts With Straight Control Limits

Viewing 8 posts - 1 through 8 (of 8 total)
• Author
Posts
• #53832

Bill Bentley
Participant

The software package I use for control charts (QI Macros) modified the p-chart algorithm so that in some circumstances it creates flat control limits for p-charts. The circumstance is if the sample sizes don’t vary a lot compared to the defect counts.

Does anyone else do this? Can someone explain the rationale for it? I’ve always used p-charts with skyline (jagged) control limits.

0
#191598

Mikel
Member

Do you have a clue how the limits are calculated?

Go to the dufus corner with Syed.

0
#191604

MBBinWI
Participant

They’re still jagged – with a step-change between sample sets of zero. When you can figure out and respond with what that means, you can get out of the dufus corner (and will be on your way to understanding control charts, not just blindly creating them).

0
#191681

Savage
Participant

Before software was so commonly used, a “rule of thumb” was to not change the limits unless the percent change in the denominator changed by more than 25% from the average denominator.

Some people make the limits flat so that they are easy to read. The problem with flat limits is that you may make more errors with interpreting the chart. (Not detecting a change when you should have.)

0
#191737

Joel Smith
Participant

If sample sizes are unequal, you should have varying limits (skyline). The prior practice of keeping them constant if the sample sizes don’t vary more than a certain amount was to make things easier for people creating a chart by hand. Since you’re using software, this simplification is not necessary.

0
#191748

Ken Feldman
Participant

Another approach used in the good ole days was to randomly sample an equal sample size from the varying data sets. For example, if the sample sizes were 50, 60, 65, 55, 45, 60 you might use the 45 as a base and then randomly select 45 from the other sample groups to keep them all constant. Simple workaround but fraught with potential problems. As was mentioned above, we did work arounds to make it simple by hand. With all the software today, why not just do it as intended and stop worrying whether the control limits are jagged or not. The “skyline” view is way cooler to look at than the straight lines. Also, possibly investigate whether you should be using attribute data rather than monitoring some underlying variable data. For example, if you are tracking proportion late possibly you should convert to actual number of minutes and then start using a variable chart such as xbar/r/s or IMr. Wheeler has also suggested that under the right conditions, you might use the proportions on an IMr chart. Lots to think about.

0
#191781

Mike Carnell
Participant

If you are going to lie to yourself about sample size changes then just plot it on a np chart. Beyond that this is a really dumb thing to do. This may be a news flash but the point of a control chart is not to control the control limits.

Just my opinion

0
#252202

Robert Tipton
Participant

Hi –

The General Rule of Thumb is +/- 25% of the Sample Size.

Example: Average Sample Size of 100 +/- 25%  =  +/-25,  it is OK to use the same Control Limits.

If the Sample Size is <=74, or >=126, then use Different Control Limits.

These different control limits are sometimes referred to as a Manhattan Skyline.

0
Viewing 8 posts - 1 through 8 (of 8 total)

You must be logged in to reply to this topic.