posted 24 July 2001 12:18 PM              

For exemple: someone working in a laboratory has to weigh fluids. He takes sample sizes of 4. This person is a bit awkward and now and then drops a bottle with fluid. So the sample size drops to 3. At the end of the measurements he wants to know the Cpm of the weight. If he uses al the measurements the samples of 4 have more effect than the samples of 3. What should he do ?
When he uses an overall sample size of 3 he has made to much measurements and he wants to use them all. (because for example he had used an expensive method for weighing)

posted 29 July 2001 11:08 AM               

quote:

---

Originally posted by student without holiday:
As I read in this forum before the Cpm-index uses every measurement and not only the means of a sample.
But when you take samples of different sizes during measuring the Cpm-index gets more influenced by the values of the larger sample sizes.

For exemple: someone working in a laboratory has to weigh fluids. He takes sample sizes of 4. This person is a bit awkward and now and then drops a bottle with fluid. So the sample size drops to 3. At the end of the measurements he wants to know the Cpm of the weight. If he uses al the measurements the samples of 4 have more effect than the samples of 3. What should he do ?
When he uses an overall sample size of 3 he has made to much measurements and he wants to use them all. (because for example he had used an expensive method for weighing)

greetings

---

Let me tell you what I think:

Before calculating Cpm index, a Control Chart must be created from the data collected. If the process is stable (the weighing process variation is under common cause of variation only), then it is apropriate to think about capability indices.

Each and every datum belongs to the population depicted in the Control Chart (remember: only common causes), therefore
the size of the sample has NO "effect" in the sense you posted.

My sugestion (strictly referredn to the problem that you stated) is that you do not need to worry about variations in the size of the samples for Cpm calculations. Such variation must be taken into account when constructing control charts.

Hope this helps.