sampling to track yield improvments

[#1]

I am looking to validate impact of process changes on yield at the step. My proposal was to track yield for 1000 samples surrounding the change, 500 pre & 500 post. Are there any benefits to separating the 500 units in each group into 5 sequential groups of 100 each and analyzing these as subgroups? Does this answer change based on the baseline yield (mid range ~60% vs high-95% + yields)

[#2 · Accepted solution]

>>Are there any benefits to separating the 500 units in each group into 5 sequential groups of 100 each and analyzing these as subgroups?

Potentially, yes. If the difference between before and after is large, you may be able to stop sampling early (assuming each subgroup is representative of the population).

>>Does this answer change based on the baseline yield (mid range ~60% vs high-95% + yields)

Yes. Power and Sample Size vary depending on how close to 50% your area of interest is. You can experiment with this with the Sample Size Explorers starting in JMP 16. A relevant location for your question is DOE...Sample Size Explorers...Power...Power for One Sample Proportion.