Capability Analysis (using Histogram) for Subgrouped Data

[rdanis]

Hello every one,
I took 50 subgroups data with sample size = 6. So totally there are 300 data.
Is it possible to run capability analysis by using histogram with
subgroupped data? If possible, how to do that? Should I do it in the
same way with sample size = 1?

Thank you.

 

[Stijloor]

A Quick Bump!

Can someone help?

Thank you very much!!

 

[Bev D]

Use all individual data points for Ppk calculations and within subgroup SD for Cpk


If you are really interested in your actual capability plot your subgroups in time series. I choose a multi vari chart that shows all data points vs th rspec limits. An alternate is to plot the subgroups in a control chart AND a histogram of the individual values but I find this to be more complex to interpret as your data may not be subgrouped rationally or your process may not be homogenous and the control chart may be misleading.


If you post your data I can create a both and explain the differences

 

[Darius]

I took 50 subgroups data with sample size = 6. So totally there are 300 data.
Is it possible to run capability analysis by using histogram with
subgroupped data? If possible, how to do that? Should I do it in the
same way with sample size = 1?.

Why "subgroups data with sample size = 6", subgroups MUST be rational, could be this group different from the next?, or just because 6 is a perfect number (according to the ancient greeks 6 = 1+2+3, the sum of their their common divisors, the next could be 28 =1+2+4+7+14)

The capability analysis is not done with an histogram, the histogram is just the representation of the call it "probability function", the % of data that can be expected to be in some interval. That you "can chart it the same way as sample size=1", that's right, but it's just a representation, NOT an analysis. The formulas for performance (pp/ppk) or capability (cp/cpk), IS part the real analysis. As Bev said, Multivari or Box plot across the time could be a good option.

carefull: An histogram of an average, could make the data look more controled than it really is as all the data variation is softened.

 

[rdanis]

Dear Bev,
Thank you for your kind response
and help.

I attach sample of data.

 

[rdanis]

Why "subgroups data with sample size = 6", subgroups MUST be rational, could be this group different from the next?, or just because 6 is a perfect number (according to the ancient greeks 6 = 1+2+3, the sum of their their common divisors, the next could be 28 =1+2+4+7+14)

The capability analysis is not done with an histogram, the histogram is just the representation of the call it "probability function", the % of data that can be expected to be in some interval. That you "can chart it the same way as sample size=1", that's right, but it's just a representation, NOT an analysis. The formulas for performance (pp/ppk) or capability (cp/cpk), IS part the real analysis. As Bev said, Multivari or Box plot across the time could be a good option.

carefull: An histogram of an average, could make the data look more controled than it really is as all the data variation is softened.

Dear Darius,

Is 6 really that bad? Taking 6 data is the current
condition in my work place and it has been implemented
since the ancient Greek before I joint the group.


BTW, Thank you for your enlightment.
You and this forum have gave me many useful
knowledge. And I have to keep studying so that
got no sarcasm from people like u.

 

[Darius]

Is 6 really that bad? Taking 6 data is the current
condition in my work place and it has been implemented
since the ancient Greek before I joint the group.

Since you don't have a sence of humor, I will keep it plain, the sample size MUST BE A RATIONAL SAMPLING, not just because....
ie. If in each shift 6 samples are taken, or some tools must be changed after a certain number of hours and 6 samples are taken in such period of time, if the lot of material is used in a time that 6 samples are taken...
So, a reason, not because you use sice the ancient Greeks, many SPC practitioners use 5 because many of the old books used to show that kind of example.

:truce:I thought that if you are taking to do histograms or charts of subgrouped data, to get the most, rational sampling must be on agenda.

 

[rdanis]

Re: Capability Analysis (represented in Histogram) for Subgrouped Data

Hi Darius,,,
Please do not misinterprete the last sentence,,
there is no any "grey" tendency to switch your mood,,


Just checked it,,
6 data refer to one of section in a international
IEC standard.

 

[Darius]

I rebuild your sheet a little, have fun

 

[Steve Prevette]

I do want to affirm the previous discussions of you need to run SPC and determine if the process is "in control". Else, any capability analysis is meaningless.

If there is no udnerlying basis for rational subgroups, 6 is as good as any number, or simply go x-individuals / moving range (which would be my own preference).