rntransform: Rank-transformation to normality

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

Value

Vector containing transformed variable, distributed as standard normal.

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.

Examples

Run this code

	# 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)