rntransform: Rank-transformation to normality

In GenABEL: genome-wide SNP association analysis

Description

Rank-transformation to normality of a variable or residuals from GLM analysis.

Usage

rntransform(formula,data,family=gaussian)

Arguments

formula	GLM formula for the variable to be transformed, or just the variable
data	data.frame or gwaa.data object containing the data
family	GLM family

Details

Rank-transformation to normality generates perfectly normal distribution from ANY distribution, unless many/heavy ties are present in variable (or residuals, if formula is used).

When formula is supplied, this procedure first calls ztransform, and then applies rank transformation to residuals.

Value

Vector containing transformed variable, distributed as standard normal.

Author(s)

Yurii Aulchenko

See Also

ztransform

Examples

	# uniformly distributed variable
	x <- round(runif(200)*100)
	# get 7 missing values
	x[round(runif(7,min=1,max=100))] <- NA
	# Z-transform
	y0 <- ztransform(x)
	# Rank-transform to normality
	y1 <- rntransform(x)
	# test normality of the original and transformed var
	shapiro.test(x)
	shapiro.test(y0)
	shapiro.test(y1)
	# plot histogram
	par(mfcol=c(3,1))
	hist(x)
	hist(y0)
	hist(y1)

Example output

Loading required package: MASS
Loading required package: GenABEL.data

	Shapiro-Wilk normality test

data:  x
W = 0.95545, p-value = 8.912e-06


	Shapiro-Wilk normality test

data:  y0
W = 0.95545, p-value = 8.912e-06


	Shapiro-Wilk normality test

data:  y1
W = 0.99785, p-value = 0.9963

GenABEL documentation built on May 30, 2017, 3:36 a.m.