iSixSigma

ANOVA Post Hoc

Viewing 3 posts - 1 through 3 (of 3 total)
  • Author
    Posts
  • #55334

    John Noguera
    Participant

    Hi all,

    A bit of informal research on the use of ANOVA post hoc tests in the Six Sigma community.

    Tukey is well known and the most popular ANOVA post hoc test, but I was wondering if anyone uses Dunnett’s (multiple comparison to control) or Hsu’s (multiple comparison to best)?

    Thanks!

    0
    #199638

    Robert Butler
    Participant

    All the time particularly Dunnett’s test. The other comparison method I use very frequently are custom contrast calls with Bonferroni adjustments. The latter is necessary when we want to limit the comparisons to a specific subset of two way comparisons because the other possible two way comparisons have no physical or clinical meaning. In these instances the contrasts are agreed upon before the analysis is run. This eliminates the urge to go on snark hunts for statistical significance in the event that nothing of interest turns up.

    0
    #199639

    John Noguera
    Participant

    Thanks Robert!

    What software are you using to do the custom contrasts?

    Also, while we are talking about post-hoc, what do you do with post-hoc on contingency tables? Adjusted Residuals seem to be the easiest (per Agresti) but curious as to what you prefer.

    Finally post-hoc on Kruskal-Wallis? Dunn is popular, but Dwass-Steel-Critchlow-Flinger looks to be more powerful.

    John

    0
Viewing 3 posts - 1 through 3 (of 3 total)

You must be logged in to reply to this topic.