topic Re: Proc Shewhart p-chart with 100% centerline in Statistical Procedures

Proc Shewhart p-chart with 100% centerline

deepika951 — Fri, 05 Jun 2020 14:20:44 GMT

Hi Everyone,
I was trying to plot a p-chart using PROC SHEWHART on SAS 9.4.

I wanted to show a positive shift in my process centerline to 100%. However, SAS shows me the following error:

WARNING: No variability was encountered in the data for process variable Completion.

Is there any suggestion on how should I show a 100% centerline?

Re: Proc Shewhart p-chart with 100% centerline

PaigeMiller — Fri, 05 Jun 2020 14:41:13 GMT

It would help if you showed your SAS code and a portion of your data.

Also: could you please explain how a p-chart can have a centerline at 100%? A p-chart has percent or proportion on the vertical axis, it goes from 0% to 100%, so it doesn't make sense to me to have the centerline at 100%.

Re: Proc Shewhart p-chart with 100% centerline

deepika951 — Fri, 05 Jun 2020 14:49:43 GMT

Week Beginning	Num	Den	Completion %
20/01/2019	8	8	100
27/01/2019	8	8	100
03/02/2019	20	20	100
10/02/2019	8	11	72.72727
17/02/2019	11	13	84.61538
24/02/2019	8	9	88.88889
03/03/2019	16	28	57.14286
10/03/2019	3	4	75
17/03/2019	19	20	95
24/03/2019	15	20	75
31/03/2019	4	4	100
07/04/2019	2	2	100
14/04/2019	5	5	100
21/04/2019	2	2	100
28/04/2019	5	5	100
05/05/2019	1	1	100
12/05/2019	6	6	100
19/05/2019	1	1	100
26/05/2019	2	3	66.66667
02/06/2019	2	3	66.66667
09/06/2019	4	4	100
16/06/2019	3	3	100
23/06/2019	4	4	100

My baseline was calculated using the 1st 8 points (~81%) and then a new centerline for 100% was calculated from the week 03/31/2019-05/19/2019 (using the 8 points above the centerline rule). However, SAS code is showing me error for not being able to calculate control limits when the centerline becomes 100%. I have tried to create 2 "Phases" in my program to calculate the 2 centerlines as shown below:

proc sort data=V; by _PHASE_;run;

proc shewhart data=V;
by _PHASE_;
pchart Completion*Week_Beginning/subgroupn=Den dataunit=percent
outlimits = LimV(rename=(_PHASE_=_INDEX_)) interval=day discrete
nochart;
run;

proc shewhart data=V limits=LimV;
pchart Completion*Week_Beginning/
markers discrete
subgroupn=Den
totpanels=1
dataunit=percent
nolegend
cout=red
height=6
odstitle=none
yscale=percent
readindexes= all
readphases=all
readlimits;
label Completion="Weekly Proportion %"
Week_Beginning="Week Beginning";
run;

Re: Proc Shewhart p-chart with 100% centerline

PaigeMiller — Fri, 05 Jun 2020 14:51:54 GMT

The error message is correct. Control limits cannot be calculated when the mean value is 100%. There is no variability from which you can compute upper or lower control limits.

Re: Proc Shewhart p-chart with 100% centerline

deepika951 — Fri, 05 Jun 2020 14:56:10 GMT

I understand. But is there a way to have control limits for the 1st Phase in which my centerline is not 100% and then have no limits when the centerline becomes 100%?

Re: Proc Shewhart p-chart with 100% centerline

PaigeMiller — Fri, 05 Jun 2020 14:58:45 GMT

Well, to be honest, I have never tried. So I cannot answer further.

Re: Proc Shewhart p-chart with 100% centerline

deepika951 — Fri, 05 Jun 2020 15:04:27 GMT

No problem. Thank you for your help anyway.