I have a question about statistical tests for dependent data. This will sound a bit strange/extremely naive but pls bear with me- here is my problem formulation:

I have 500 individuals, and I'm interested in statistics on groups of pairs of these individuals. In others words, I've enumerated all pairs of these individuals and separated the pairs into two groups, i.e.:

Group A: pairs of individuals with significant sharing of mutations in a particular gene X (i.e. each pair of individuals has high overlap of mutations in gene X)

Group B: pairs of individuals with no significant sharing of mutations in gene X

Now, I'd like to see whether Group A has higher expression of some gene Y (e.g. 1 is low expression, 1000 is high) compared to Group B. Is there a way to do this?

Obviously the pairs are not independent, so I'm not sure whether it is even possible to formulate the problem this way. In my case, the pairs are the quantities of interest- not the individuals. Each pair only appears in one of the groups, but since I've enumerated all possible pairs, many individuals appear in both groups.

I guess the general question is, what statistics can be performed (if any) between two groups of pairs with respect to some continuous variable? I'm pretty sure this violates the independence of all simple tests (e.g. t-test, wilcoxon, etc) so not sure how to proceed.

Would greatly appreciate any tips. Thanks!

  • $\begingroup$ How is your dependent variable (expression of gene Y) measured? At the individual or pair level? If at pair level, is it some function of what each individual in the pair has? I am just trying to better understand what the significance of pairing individuals is in terms of how it might affect the outcome. $\endgroup$ – AlexK May 7 at 1:21
  • $\begingroup$ @AlexK, thanks for your response. I should have clarified, expression of gene Y is also at the pair level (i.e. a mean expression of gene Y of individual 1 of the pair, and individual 2 of the pair). $\endgroup$ – qstats May 7 at 1:50
  • $\begingroup$ While most commonly-used paired tests look at pair-differences, presumably the ratio of expression would be of interest here (which might lead us to look at differences of pairs of logs of gene expression values). $\endgroup$ – Glen_b May 7 at 4:41
  • $\begingroup$ Are these "mean expression" values actually averages of counts? $\endgroup$ – Glen_b May 7 at 4:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.