+ Reply to Thread
Results 1 to 4 of 4

Thread: Which test to use for binary variables and concerns about multiple comparisons

  1. #1
    Points: 7, Level: 1
    Level completed: 13%, Points required for next Level: 43

    Posts
    2
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Which test to use for binary variables and concerns about multiple comparisons




    Hi all,

    I am working on a research project attempting to compare two groups of subjects' (call them A and B) brains based on MRI results.

    Let's say in my data set I have split each subject's brain into 50 regions, and then classified each region as either "healthy" or "not healthy" i.e. assigned a 1 or 0 as a binary variable.

    If I compare the total number of "not healthy" brain regions for each subject across these two groups, I see they are different (p<.05 using either two-sample t-test or a non-parametric rank-sum test) and would like to explore if a subset of particular regions is driving this group difference. Let's say Group A tends to have fewer total "healthy" regions than Group B.

    Is the appropriate test now to do a two sample chi-square test for each region? If so, wouldn't I need to correct my p-values for multiple comparisons using either Bonferroni or FDR and need a p<.001 (i.e. 0.05 / 50)?

    Is it legitimate to only look at the regions with lower average healthy % in group A and do a multiple comparisons correction across that number (which should be less than 50 presumably)?

    Thanks in advance,
    -mnd12

  2. #2
    Omega Contributor
    Points: 39,022, Level: 100
    Level completed: 0%, Points required for next Level: 0
    hlsmith's Avatar
    Location
    Not Ames, IA
    Posts
    7,069
    Thanks
    402
    Thanked 1,192 Times in 1,153 Posts

    Re: Which test to use for binary variables and concerns about multiple comparisons

    Why not primary outcome two sample chisq, uncorrected (proportions healthy).

    Then secondary outcomes, 50 regions, using two sample chisq, yes alpha level corrected.
    Stop cowardice, ban guns!

  3. #3
    Points: 7, Level: 1
    Level completed: 13%, Points required for next Level: 43

    Posts
    2
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: Which test to use for binary variables and concerns about multiple comparisons

    Quote Originally Posted by hlsmith View Post
    Why not primary outcome two sample chisq, uncorrected (proportions healthy).

    Then secondary outcomes, 50 regions, using two sample chisq, yes alpha level corrected.
    Thanks for your response!

    For the primary outcome, I have a proportion for each subject, say 20 in both group A and B--how do I use these 40 proportions into a 2 sample chi-square test? I thought a two sample chi-square was for comparing 2 proportions?

    Thanks
    mnd12

  4. #4
    Omega Contributor
    Points: 39,022, Level: 100
    Level completed: 0%, Points required for next Level: 0
    hlsmith's Avatar
    Location
    Not Ames, IA
    Posts
    7,069
    Thanks
    402
    Thanked 1,192 Times in 1,153 Posts

    Re: Which test to use for binary variables and concerns about multiple comparisons


    What is your sample size?
    Stop cowardice, ban guns!

+ 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