iSixSigma

2-sample t-test failed, can I still compare capability analysis?

Six Sigma – iSixSigma Forums General Forums General 2-sample t-test failed, can I still compare capability analysis?

Viewing 4 posts - 1 through 4 (of 4 total)
  • Author
    Posts
  • #255871

    Peteriks
    Participant

    I want to compare two categorical situations (old vs. new part in a machine) with the continuous CTQ, the y-variable.

    In order to compare the 2 situations, I found the proper distribution and compared it to two Ppk values. I saw a significant difference in performance (the new part was +-3 times better in terms of exceeding the USL and LSL).

    However, I also performed a 2-sample t-test and test for equal variance. Both of these ‘failed’ and showed no significant difference in either the mean or variance between the population of the new part and the old part. I also performed a power & sample size test to ensure I had enough data points in my samples/populations.

    Can I still use the difference between the capability analysis as evidence of the new part significantly outperforming the old part, or is this not possible since there is no significant difference between the mean and variance?

    Thanks in advance!

    • This topic was modified 6 months, 3 weeks ago by Peteriks.
    0
    #255873

    Chris Seider
    Participant

    Are you sure you did the comparison correctly? Did you wait until the mean and/or standard deviation looked at a stable process reading before comparing the means or sd?

    0
    #255879

    Peteriks
    Participant

    Hi Chris, thank you for the reply. I think you might be right. I checked the power&sample size analysis again and saw that the sample size for the ‘old’ part is sufficient, but the sample size for the ‘new’ part is too little (70 instead of 200). Is this prove that the SD/mean are not stable yet, or am I wrong?

    I will postpone further analysis until stabilisation. I reckon if it’s not stable I also cannot trust the capability analysis yet, right?

    PS. I analysed it for another CTQ as well, where the sample size for the ‘new’ part is the same, however, here a meaningful difference in mean/SD and ppk was noticed, so I do believe I can use that data even though the sample size is not large enough yet, right?

    Thanks in advance and thanks for the initial reply!

    0
    #255915

    kathilee
    Participant

    Thanks for helping with this!

    0
Viewing 4 posts - 1 through 4 (of 4 total)

You must be logged in to reply to this topic.