The capability index doesn´t depend on the lot size, of course the estimates of mean and standard deviation could be better when more data is added to the study but the calculus is just the same.
I tink your cuestion is about the confidence of the index calculus.
https://www.qualitydigest.com/may00/html/lastword.html
and the calculus of the index is just the same as with subgrouped data.
Cpk is the minimum of (USL- Mean, Mean - LSL)/(3 sigma)
Where sigma in individual control charts is calculated
Sigma = sum(MR)/(lot_size - 1) /1.128
that is the same sigma used to calculate the control limits
(broken link removed)
The problem that I found is that when there is autocorrelation is the data, you have to add the autocorrelation factor to the Sigma (multiply by the factor), this happends most frecuently when there is automatic data sampling or in batch processing.
I hope this may help.
:thedeal:
