Assessing whether differences between datasets are significant I am unsure which statistical test I should use to test whether two datasets are significantly different. The data looks something like this:
Group   Dataset 1   Dataset 2
A   5   6
B   1   7
C   2   10
D   1   15
E   4   11
F   1   5
G   2   17
H   2   10
I   3   19
J   1   6
K   14  67
L   13  38
M   3   16
N   14  59
O   14  33
P   0   4
Q   10  28
R   0   6
S   4   9
T   3   5
U   3   13
V   0   13

I am hoping that, although one dataset has a larger sum, the proportions between groups mean that they are not significantly different. 
Does anybody have any helpful advice or suggestions of a suitable test? Thanks in advance
 A: It sounds like you're after a test of homogeneity (i.e. whether the proportions in each classification are the same across the two groups of genes).
The usual way to do this is a via a chi-squared test.
Below are the results from performing (in R) the chi-squared test against the usual chi-squared distribution, and by simulating the distribution of the test statistic (conditional on the margins). 
The common rule of requiring all expected values greater than 5 is usually too stringent, and it seems to have been the case here; the simulated p-value is essentially the same.
Using chi-squared distribution:
> chisq.test(genecl[,2:3])

    Pearson's Chi-squared test

data:  genecl[, 2:3]
X-squared = 22.4867, df = 21, p-value = 0.3719

Warning message:
In chisq.test(genecl[, 2:3]) : Chi-squared approximation may be incorrect

Simulated:
> chisq.test(genecl[,2:3],simulate=TRUE,B=100000)

    Pearson's Chi-squared test with simulated p-value (based on 1e+05 replicates)

data:  genecl[, 2:3]
X-squared = 22.4867, df = NA, p-value = 0.3704

In either case the proportions in each classification are not sufficiently different to indicate a difference in the distribution across classifications.
