# Subgroup size - Capability Analysis

#### Deano P

##### Registered
Hi all. If you have a scenario, where you can only take a sample at the beginning and end of the process (the nature of the product only allows this) and this goes on for a duration for a month and lets say you get x30 data readings. would this then be classed as a subgroup size of 1? even though you are getting x2 data points from that run or would you class it as a subgroup size of 15

#### Bev D

##### Heretical Statistician
Super Moderator
I would start with a subgroup size of 1. But there is a scenario where the subgroup size may be 2 (the 2 measurements at teh beginning and end of each run). It depends on how homogenous or non-homogenous your process is.
The reason lies in the statistical basis of rational subgrouping. See this article on Rational Subgrouping by Donald Wheeler.

In the mean time the best approach is to try a control chart based on a subgroup of 1 and a chart with a subgroup of 2. Which one demonstrates that the between subgroup variation is very small compared to the within subgroup variation. A homogeneous process has very little between subgroup variation…of course you may not have a stable process so…you can post your results and data here and we can take a look..

#### Deano P

##### Registered
Hi Bev, Thanks for your response. There are outliners on the control chart for a subgroup size of 1 and 2 (more than 3 standard deviations for the centre line. Like i said in the original email, the measurements represent the diameter of the product, the measurements are taken from the beginning of the run and the end of the Run. The tolerance is very tight also, with leaves no room for variation within the process. i am talking about micron tolerance's.

• As there is evidence of variation from the control charts, what can you suggest
• Do you think you should stick to subgroup 1 or 2?
• When doing the capability analysis for CP,CPK,PP,PPK should you use between/within (as there's subgroups) or normal, non normal?
I can see the is a slight difference in CP and CPK when you use the between/within and normal (CP,CPK slightly lower using between/within)