Home Forums PATN and Pattern Analysis PATN Discussion paired anosim results

  • This topic is empty.
Viewing 2 posts - 1 through 2 (of 2 total)
  • Author
    Posts
  • #418
    donald
    Member

    Dear Lee

    I have assumed that the paired anosim evaluations allow one to see which groups are needed in the final groupings.
    The overall ANOSIM simply tests whether the grouping chosen is better than random, amongst a set of groups of the same size. But if any group is strongly differentiated from the rest the overal f-ratio may well be significantly different from 1, regardless of how justified the decision to choose 19 rather than 10 groups is.

    I have presumed that you have put in the paired ANOSIM tests to get over the fact that the overall f-ratio test is a very blunt instrument.

    I was hoping you might comment.

    Margaret

    #491
    lee
    Keymaster

    Hi Margaret

    Yes, the overall ANOSIM is rather ‘blunt’. I have never run a classification where the resulting PATN-produced groups were not significantly different from random. The test was designed more for a-priori groups.

    The pair-wise ANOSIM is doing the same, but limiting itself to taking each pair of defined groups independently from all other groups. If you had two groups, overall and pair-wise should be the same.

    The pair-wise test could be used to identify if the number of groups is reasonable. I would however think that PATN generally does such a good job that you would find that you would have to generate many more groups than convenient to get down to a point where discrimination maybe lacking. Personally, two-way tables (and box and whisker plots) will give you a simpler way of understanding the heterogeneity.

    So, while pair-wise ANOSIM is focusing on a smaller (simpler) domain, I still think it is somewhat tangential.

    Lee

Viewing 2 posts - 1 through 2 (of 2 total)
  • You must be logged in to reply to this topic.