# Determining Stastically significant difference in sizes of descrete groups

I have data from a experiement involving DNA, where a "random" string of DNA is created.

This would look like [ [A,C,T,G],[A,C,T,G],[A,C,T,G]] where the string is any of the combinations of those four letters. A set of example strings is: AAA, CTG, GTA . I would also like to look at anything from 1 to 6 letter combinations however.

I want to test if my experiment resulted in a statistically significant higher frequency for a given string (I am expecting a higher instance of G's). I have a set of arrays for all of the different groups so I have effectively (1 to 6)^4 different groups for each test.

I was thinking of using a ANOVA test, but I have a computer background and the researcher I am helping doesnt have any better stats skills than I do. Can anyone give a suggestion and please post some reasources for a good way to do this?

• Have you considered a chi-squared goodness of fit test? Sep 21 at 4:01
• I have looked at chi-squared tests, the examples of their use seem to check if the distribution overall is biased. I want to see if a specific category is being biased towards. Roughly speaking I think I should be using a variance term at some point to account for the fact that other categories may have above/below average values as well. Sep 22 at 17:07