DOE
Six Sigma – iSixSigma › Forums › Old Forums › General › DOE
- This topic has 13 replies, 8 voices, and was last updated 13 years, 1 month ago by
Burichi.
-
AuthorPosts
-
March 18, 2009 at 10:17 am #52051
BurichiParticipant@BurichiInclude @Burichi in your post and this person will
be notified via email.hi,anybody who could help me figure out what to do? i have ran an experiment… 2^2 factorial design… my problem is my response variable is an attribute/proportion data. is factorial design applicable for this case? coz i was thinking that factorial design is only applicable when you have variables type of response.thanks in advance!
0March 18, 2009 at 11:04 am #182485It’s fine, you just needs lots of data – go look at sample size for a two
proportion test and make sure you have it covered.0March 18, 2009 at 4:30 pm #182492You can run DOE for attribute data. all you have to do is collect 10 or 20 samples (for ex) for each run and check the no. of fails or pass.
Ashky0March 18, 2009 at 4:45 pm #182493Your advice of 10 or 20 is dangerous. Sample size depends on the p value, how big of a difference you
need to detect and desired confidence. You sample size is likely
much bigger than your suggestion.0March 18, 2009 at 4:51 pm #182494
MBBinWIParticipant@MBBinWIInclude @MBBinWI in your post and this person will
be notified via email.and if you end up not having good replicates, you can always fall back on Chi Sq. just using counts for each condition.
0March 18, 2009 at 5:25 pm #182496Stan,
got your note…made contacts in/out.0March 18, 2009 at 7:58 pm #182498Your advice of 10 or 20 is dangerous. Sample size depends on the p value, how big of a difference you
need to detect and desired confidence. You sample size is likely
much bigger than your suggestion.0March 19, 2009 at 5:24 am #182507Burichi,
Why didn’t you ask this question before you ran the experiment? I have performed factorial experiments in the SMT industry where the quantity of solder balls was the response. (Count data). You should always anticipate how you will analyze your data before you collect it.
HACL0March 19, 2009 at 8:02 am #182508
Bower ChielParticipant@Bower-ChielInclude @Bower-Chiel in your post and this person will
be notified via email.Hi BurichiIn the book Process Quality Control, (subtitled Troubleshooting and Intrepretation of Data) by Ellis Ott et al, the numbers tested for each combination of factor levels range in the examples he gives range from around 40 upwards. The book uses Analysis of Means (ANOM) rather than Analysis of Variance (ANOVA) but has lots of good examples in it. I have the third edition and it comes with a CD with software for ANOM.Best WishesBower Chiel
0March 19, 2009 at 11:22 am #182511You did not share the nature of your data.. some attribute data can be stretched into continuous data if it is correctly setup.
Did you have simply pass or fail or did you have degrees of goodness?0March 19, 2009 at 1:22 pm #182516
MBBinWIParticipant@MBBinWIInclude @MBBinWI in your post and this person will
be notified via email.Oftentimes there was actual continuous data gathered, then it was evaluated and a judgement of “good” or “bad” was made. See if there was actual continuous data to begin with. Surprisingly, I’ve seen organizations “flush” continuous data, but retain forever pass/fail.
0March 23, 2009 at 2:39 am #182630
BurichiParticipant@BurichiInclude @Burichi in your post and this person will
be notified via email.why use two proportions to get the sample size? should i use one proportion?
0March 23, 2009 at 2:47 am #182631
BurichiParticipant@BurichiInclude @Burichi in your post and this person will
be notified via email.so what did you use in your DOE? what kind of experimental design?
0March 23, 2009 at 2:51 am #182632
BurichiParticipant@BurichiInclude @Burichi in your post and this person will
be notified via email.for each legs i ran a few samples… i can compute for the proportion for each leg. can i use a factorial design for this?
0 -
AuthorPosts
The forum ‘General’ is closed to new topics and replies.