+ Reply to Thread
Results 1 to 5 of 5

Thread: Ecology : re-group samples based on distance measure

  1. #1
    Points: 1,045, Level: 17
    Level completed: 45%, Points required for next Level: 55

    Posts
    13
    Thanks
    3
    Thanked 0 Times in 0 Posts

    Ecology : re-group samples based on distance measure




    Hello,

    Good morning TalkStat Forum, hedre is my question of the day

    I have an ecological matrix of samples (rows) X species abundance (columns).

    Originally the samples were organized in 3 groups (let's start by assuming they are independant) : one control group, 2 treatment groups. Analyses did not show any significant differences between the groups : no clustering in the dendrogram, no clustering in NMS or PCA, no significant difference using MRPP test (Bray-Curtis distance).

    However I noticed that, while the majority of the test samples were very close to the control sampes, a subset of the test samples was quite far (NMS observations). I then made the hypothesis that my test samples actually comprised 2 populations : one identical to the controls, and one that was different. I re-analyzed the cluster dendrogram and noticed that indeed some test samples (from both test groups) clustered separately from any control samples. I put these samples in a new group, and now this new group is significantly different from the control group (MRPP), and clusters separately from the control group on NMS and PCA.

    I am quite happy with the result. But all this is only based on a clustering cut off.

    So here is my question (thank you for your patience) : is there a way to optimize the group separation ? Is there another test I can do to make sure that my new group is optimally separated from my control group (Bray-Curtis disrance ? UniFrac measure ?) ? For the moment it seems a little arbitrary (I only have 1 test to justify this separation), and I am worried about false negative (leaving interesting samples out of my new group) or false positive.

    Any suggestions ?
    Sheldon : Well, there’s always the possibility that a trash can spontaneously formed around the letter, but Occam’s Razor would suggest that someone threw it out.

  2. #2
    Super Moderator
    Points: 31,766, Level: 100
    Level completed: 0%, Points required for next Level: 0
    bugman's Avatar
    Posts
    2,255
    Thanks
    290
    Thanked 324 Times in 265 Posts

    Re: Ecology : re-group samples based on distance measure

    If you are using PRIMER with the PERMANOVA add in you can look into Canonical Analysis of Princioal Components (CAP) which allows optimum sepration of a-priori groups. However, what you described sounds a bit to me like dat snooping and is not very good practice. Anyway, I wont debate that. Check out CAP. It has some interesting features.
    The earth is round: P<0.05

  3. The Following User Says Thank You to bugman For This Useful Post:

    Koblenz (07-02-2013)

  4. #3
    Points: 1,045, Level: 17
    Level completed: 45%, Points required for next Level: 55

    Posts
    13
    Thanks
    3
    Thanked 0 Times in 0 Posts

    Re: Ecology : re-group samples based on distance measure

    Hello,

    Thank you for your answer. I don't have Primer-e, I use the softwares PCord, R and JMP. I will try the demo version and test the PERMANOVA.

    I know it is generally not a good idea to re-arrange groups a posteriori, but in this case it is very likely that the test subjects are divided in two subsets, one ressembling the control subjects, the other one with different characteristics. And the main answer I am trying to obtain here is : what makes this subset of subjects different, when the test group as a whole is not different from the controls ? But for that I need to know what subjects can objectively be considered significantly different from the controls.
    Sheldon : Well, there’s always the possibility that a trash can spontaneously formed around the letter, but Occam’s Razor would suggest that someone threw it out.

  5. #4
    Super Moderator
    Points: 31,766, Level: 100
    Level completed: 0%, Points required for next Level: 0
    bugman's Avatar
    Posts
    2,255
    Thanks
    290
    Thanked 324 Times in 265 Posts

    Re: Ecology : re-group samples based on distance measure

    Your reasoning sounds logical and it looks like you have it sorted. Good luck!

    Just doing a quick seacrh, I think CAP might be available in the vegan package for R.
    The earth is round: P<0.05

  6. #5
    Points: 1,045, Level: 17
    Level completed: 45%, Points required for next Level: 55

    Posts
    13
    Thanks
    3
    Thanked 0 Times in 0 Posts

    Re: Ecology : re-group samples based on distance measure


    Yes, I know what I want. I'm just looking for the best way to do it

    I'll look into the options of the vegan package.
    Sheldon : Well, there’s always the possibility that a trash can spontaneously formed around the letter, but Occam’s Razor would suggest that someone threw it out.

+ Reply to Thread

           




Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats