10-17-2014 03:23 PM
I am working with a manufacturing process data and cannot figure out how to control the inset calculations. For example: Lets say I have 80 data results that I am graphing using the Proc Shewhart procedure with an irchart. But between data point 40 and 41 I split the control limits due to a shift in the mean. So I have my control limits for data points 1-40 and a second set of control limits for 41-80. Specification limits for all data is the same. I then use the inset statement to provide a CPK point estimate along with the 95% confidence interval points. However, I want the inset statement to only reflect on the data just within the second set of control limits.
proc shewhart history=data2 limits=stats;
irchart test*lot_count='*' / outlabel=(lot_count) nochart2 vref=vref ciindices lsl=&lsl usl=&usl readphases=all
readindexes=all phaselegend vreflabpos=2 zerostd tests=1 to 4 testlabel1=' ' testlabel2=' ' clipfactor=2 totpanels=1;
inset stddev cpklcl='Cpk 95% Lower' cpk='Cpk' cpkucl='Cpk 95% Upper' / pos=sw format=6.2 opacity=0;
Where data2 contains the two different _PHASE_ values, and stats data set contains the two sets of control limits with _INDEX_ values to be matched with _PHASE_ values.
And also, how can I do this for a histogram in proc capability. I want the histogram to include all data points in the chart, but I only want the ppk calculations to be based off only the 2nd half of the data.
10-18-2014 06:46 AM
I don't have an easy answer. Makes sense that if you have a phased control chart, you would want phased Cpk (or in your case Cpk for just the most recent phase).
Typically I haven't used the graphics from Shewhart much. I've instead used the proc to compute the data I want (process limits, Cpk, etc) and output them to a dataset, then used GTL to create a custom chart. It's more work, but gives more control over the final output. So with that approach, could just run Shewhart a second time using only data from the most recent phase to to get the desired Cpk.