snpprior

(Beta) Binomial prior for number of SNPs in a model ' ' A binomial

snpprior(x = 0:10, n, expected, overdispersion = 1, pi0 = NA,
  truncate = NA, overdispersion.warning = TRUE)

Arguments

x	number of SNPs in a model (defaults to 1:length(groups), ie returns a vector)
n	total number of SNPs or SNP groups available
expected	expected number of SNPs in a model
overdispersion	overdispersion parameter. Setting this to 1 gives a binomial prior. Values < 1 are nonsensical: if you really believe the prior should be underdispersed relative to a binomial distribution, consider using a hypergeometric prior.
pi0	prior probability that no SNP is associated
truncate	optional, if supplied priors will be adjusted so models with x>truncate have prior 0
overdispersion.warning	by default, prior distributions should be binomial or beta-binomial (overdispersed). If you give an overdispersion <1, snpprior will stop with an error. Set overdispersion.warning=FALSE to override this.

Value

prior probabilities as a numeric vector

Examples

n<-100 # 100 SNPs in region
x <- 1:10 # consider prior for up to 10 causal SNPs
xbar <- 3 # expect around 3 causal

## a binomial prior
y <- snpprior(x, n, xbar)
plot(x, y, type="h")

## is equivalent to
y1.0 <- snpprior(x, n, xbar, overdispersion=1.0)
points(x, y1.0, col="red")

##larger values of overdispersion change the distribution:
y1.1 <- snpprior(x, n, xbar, overdispersion=1.1)
y1.5 <- snpprior(x, n, xbar, overdispersion=1.5)
y2.0 <- snpprior(x, n, xbar, overdispersion=2.0)
points(x, y1.1, col="orange")
points(x, y1.5, col="pink")
points(x, y2.0, col="green")