I have observations from two data-sets outlining the number of nucleotide differences. Observations from the first data-set are from actual data and the second data-set contains simulated observations.

    |C-A | C-T| G-A|
dat1|10  | 5  |  20|
dat2| 15 | 14 |  14|

I would like to test if there are differences between, say C-A changes in the first data-set and in the second data-set for each type of change (C-A, C-T and G-A)

for example, Null hypothesis: There are no differences between the number of C-A changes between the actual data and the simulated data.

Which is an appropriate statistical test to allow me to reject or accept this hypothesis?

  • 1
    $\begingroup$ A $\chi^2$ test for contingency tables seems appropriate to me. $\endgroup$
    – Andy W
    Jan 23, 2015 at 12:54
  • $\begingroup$ My concern is that the two data-sets are not of the same size. or does that not matter? $\endgroup$
    – eastafri
    Jan 23, 2015 at 13:06
  • 2
    $\begingroup$ That makes no difference @eastafri. The expected values are calculated based on the marginals of the table, which take the differing sample sizes into account. (You can transpose the table and it will still result in the same test statistic.) $\endgroup$
    – Andy W
    Jan 23, 2015 at 13:13


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.