p.combine: Combine two p-values

In gerstung-lab/deepSNV: Detection of subclonal SNVs in deep sequencing data.

Combine two p-values

Description

This function combines two P-values into a single one using a statistic defined by method. "fisher" uses the product of the two, in this case the logarithm of the product is χ^2_4 distributed. If the method = "max", the resulting P-value is \max\{P_1,P_2\}^2. For method = "average" the mean is used, yielding a P-value of 2 x^2 if x=(P_1+P_2)/2 < .5 and 1-2 x^2 otherwise. "negfisher" is the negative of Fisher's method using $1-F(1-P_1, 1-P_2)$, where $F$ is the combination function of Fisher's method; for small $P_1,P_2$, the result is very similar to method="average". Fisher's method behaves a bit like a logical AND of the joint null-hypothesis, whereas negative Fisher is like an OR.

Usage

p.combine(p1, p2, method = c("fisher", "max", "average", "prod", "negfisher"))

Arguments

p1	P-value 1
p2	P-value 2
method	One of "fisher" (default), "max" or "average"

Value

p-values

Author(s)

Moritz Gerstung

Examples

p1 <- runif(1000)
p2 <- runif(1000)
hist(p1)
p.avg = p.combine(p1,p2, method="average")
hist(p.avg)
p.fish = p.combine(p1,p2, method="fisher")
hist(p.fish)
p.max = p.combine(p1,p2, method="max")
hist(p.max)
pairs(data.frame(p1,p2,p.fish,p.max,p.avg))

gerstung-lab/deepSNV documentation built on June 3, 2022, 3:05 p.m.