| T O P I C R E V I E W |
| Josh |
Posted - 11 Sep 2008 : 14:23:17
I have an application that's creating XbarR SPCBatchVariableControlCharts for user entered data. Everything works fine, except for when the user only enters their data 1 sample at a time (which causes the data to be pulled for the chart with only 1 sample per subgroup). Under the condition, the CPK value is not calculated and the Control Limits are layed over the mean (or very close to it).
Is there something that I need to change to get my charts to work with subgroups of 1?
Also, my application obviously doesn't force the users to enter their data to a fixed subgroup size (although it does limit to 5); could this cause any problems with the CPK or control limit calculations? Say for instance they enter their data in the following subgroup sizes: 5, 5, 1, 2, 5? The chart is being initialized with the maximum subgroup size that the data contains (5 in this example).
Thanks. |
| 3 L A T E S T R E P L I E S (Newest First) |
| Josh |
Posted - 11 Sep 2008 : 16:53:55
Thanks for the reply. I'll look into the I-R charts for single sampled data.
|
| quinncurtis |
Posted - 11 Sep 2008 : 15:31:33
1. First, having a single sample in a sample subgroup is completely incompatible with an X-Bar R Chart. You MUST use the Individual Range (I-R) chart if you have a single sample per subgroup. The X-Bar R, X-Bar Sigma, and Median-Range charts all require more than one sample per subgroup.
2. Having the same value for each sample in a sample subgroup is just as bad. The process in that case would violate the requirement of normality within each sample subgroup. That makes all calculations in the X-Bar R chart invalid.
Read any of the textbooks we reference in the manual for detailed information about the sample requirements for SPC.
Item 2 should be impossible in a real-world process (or it is one that cannot be monitored and improved using SPC), and it seems you came to that test case because of Item 1, which is also invalid, because X-Bar R charts require more than one sample per sample subgroup.
Charts that use Variable Sample subgroup sizes are DIFFERENT than those that use a fixed sample subgroups size. You CANNOT use a variable sample subgroup size with our standard FIXED sample subgroup size charts, which is all that we had in revisions of the software prior to Rev. 1.7. Instead, you must use the special charts we added in Revision 1.7 of the software. This is documented in the manual. There is no such thing as an X-Bar R, variable samples per subgroup chart; it is statistically invalid. The only variable control chart that permits a variable number of samples per subgroup is the X-Bar Sigma chart (you must use our special variant of that chart). A chart that has a variable number of samples per subgroup will have limit lines that vary from sample interval to sample interval. Even with a chart that supports a variable number of samples per subgroup, it is still invalid to have one sample per subgroup.
|
| Josh |
Posted - 11 Sep 2008 : 15:06:25
FYI, I'm using version 1.7.1.4.
I just tried altering my data population routine to prevent a subgroup size of 1, and discovered that the problem appears to actually be related to not having any deviation in the subgroup (range of 0).
Let's say that I enter all my data in subgroups of 5, but each sample is the same within each subgroup, the CPK will be calculated as 0 and the UCL, LCL, and XBar will overlay each other. This happens even if the samples for each subgroup are a different value, but the same value within each subgroup.
ie:
samples(0) = 2, 2, 2, 2, 2
samples(1) = 3, 3, 3, 3, 3
samples(2) = 1.5, 1.5, 1.5, 1.5, 1.5 |
|
|
