- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
when I run the below code to perform the special cause test, I receive the warning shown below. Could you please advise on how to fix that?
proc shewhart data=tmp10;
pchart yes_answer*month /Markers subgroupn = n_per_group ODSTITLE="P-chart with Tests for Special Causes"
tests = 1 to 4
testnmethod = standardize
table
tablelegend
nohlabel;
LABEL yes_answer ="Prop of Pts Receiving IV Abx";
run;
WARNING: Asymmetric control limits encountered for yes_answer for at least one subgroup.
WARNING: It is assumed that the difference between the upper limit and center line represents three standard errors.
Thank you!
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Asymmetric control limits often happen for a p-chart, because the lower control limit is truncated at 0. The warning says "It is assumed that the difference between the upper limit and center line represents three standard errors." I would think that's a fair assumption, I don't see anywhere in your code that asked for different control limits. On the chart, it shows the upper control limit as a 3-sigma limit, right? I don't think I'd worry about this warning.
Next up: Rick Wicklin presents Ten Tips for Effective Statistical Graphics (with SAS code) on Wednesday March 26.
Register now at https://www.basug.org/events.
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Thank you for your answer, I have attached the chart, which indicates that the UCL is set at a 3-sigma limit. However, the LCL does not show a corresponding 3-sigma limit. I think this discrepancy is due to varying sample sizes and the proportion being very high for each subgroup, with the UCL approaching one. (in my case LCL doesn't truncate at zero)
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Agree, you've got 3-sigma limits. But in your case the UCL is truncated by the maximum value, 1. I wouldn't worry about the warning message. Looking at tests 1-4, I think they should be fine.
Next up: Rick Wicklin presents Ten Tips for Effective Statistical Graphics (with SAS code) on Wednesday March 26.
Register now at https://www.basug.org/events.