compare_af.r

compare allele frequencies

# simulation
n <- 10000
a <- rbinom(n, 2, 0.5)
b <- rbinom(n, 2, 0.49)

# perform fisher's exact test to obtain odds ratio for two allele frequencies being different
cont <- matrix(
	c(sum(a==0) * 2 + sum(a==1), sum(a==2) * 2 + sum(a==1), sum(b==0) * 2 + sum(b==1), sum(b==2) * 2 + sum(b==1)), 2, 2)

fisher.test(cont)

# can do this approximately with just allele frequency and sample size

# make sure that the effect allele is the same for the two populations
compare_af <- function(eaf_pop1, eaf_pop2, n_pop1, n_pop2)
{
	contingency <- matrix(
		c(2 * eaf_pop1 * n_pop1, 2 * (1-eaf_pop1) * n_pop1, 2 * eaf_pop2 * n_pop2, 2 * (1-eaf_pop2) * n_pop2), 2, 2)
	x <- fisher.test(contingency)
	return(x)
}

# example
compare_af(0.2, 0.19, 10000, 10000)

# get the odds ratio using 
x <- compare_af(0.2, 0.19, 10000, 10000)
x$estimate

# probably best to use the log odds ratio for any further analysis as it is symmetrical
log(x$estimate)