Sample size calculation

Tahirawan77

Involved In Discussions
Hi,

I want to calculate the 'sample size' requirement for a before / after process improvement. The defect rate before improvement is 13% and the defect rate after process improvement is 3%. So the goal is to prove that the defect rate has been reduced from 13% to 3% and have enough samples tested to prove the statement. I would like to demonstrate a Power value of 95%. Which test should I use. Is it sample size calculation for 1 proportion or 2 proportion test?

I have used the 2 proportion test and got a sample size of 133 (see attachment). Is it the correct answer?

Thanks
 

Attachments

  • sample size.JPG
    sample size.JPG
    50 KB · Views: 57
Last edited:

Bev D

Heretical Statistician
Leader
Super Moderator
Have you already taken the after samples?
Do you have time series data of the before defect rates? There is a much more powerful approach to demonstrating that the ‘after‘ process is better than the ‘before’ process…
 

Tahirawan77

Involved In Discussions
Yes, I already have taken the 'after' samples. And if you mean 'Time series data' for 'before' then yes i also have this data. What is the other powerful technique to demonstrate that 'after' process is improved? I only have the binary data (pass/fail) for both 'before' and 'after' the process.

Thanks
 

Bev D

Heretical Statistician
Leader
Super Moderator
Plot the before data on a control chart. Then look at the after sample result in comparison to the before data. Is the after result lower than the lowest before data? Is the after result lower than the before control limits? If so then the after is better than the before (its called the “bloody obvious“ statistical test. Actually it’s probability rather than inferential statistics.). If it’s within the control limits then the amount of improvement may be either simple chance OR the shift is too small to detect with a single ‘after’ sample. In this case you will multiple after samples to detect a real sustained improvement…

I recommend that you read “On Probability as the Basis for Action” by Deming. A simple internet search will yield several sites for the article that are free.
 
Top Bottom