testPhenotype: Test the Phenotype of Interest for Association with...

In andreyshabalin/ramwas: Fast Methylome-Wide Association Study Pipeline for Enrichment Platforms

Description

An internal, function for fast association testing. It tests the phenotype of interest for association with methylation coverage (columns of the data parameter).

Usage

testPhenotype(phenotype, data1, cvrtqr)

Arguments

phenotype	Vector with phenotype. Can be numerical, character, or factor vector.
data1	Matrix with data (normalized coverage), one variable (CpG) per column.
cvrtqr	Orthonormalized covariates, one covariate per column. See orthonormalizeCovariates.

Details

The testing is performed using matrix operations and C/C++ code, emplying an approach similar to that in MatrixEQTL.

Value

If the phenotype is numerical, the output is a list with

correlation	Correlations between residualized phenotype and data columns.
tstat	Corresponding T-statistics
pvalue	Corresponding P-values
nVarTested	Always 1
dfFull	Number of degrees of freedom of the T-test

If the phenotype is a factor (or character)

Rsquared	R-squared for the residualized ANOVA F-test.
Fstat	Corresponding F-test
pvalue	Corresponding P-values
nVarTested	First number of degrees of freedom for the F-test. Equal to the number of factor levels reduced by 1
dfFull	Second number of degrees of freedom for the F-test.

Note

This function is used in several parts of the pipeline.

Author(s)

Andrey A Shabalin [e-mail]

See Also

See vignettes: browseVignettes("ramwas").

Also check orthonormalizeCovariates.

Examples

### Generate data inputs
# Random data matrix with signal in the first column
data = matrix(runif(30*5), nrow = 30, ncol = 5)
data[,1] = data[,1] + rep(0:2, each = 10)

# Two random covariates
cvrt = matrix(runif(2*30), nrow = 30, ncol = 2)
cvrtqr = orthonormalizeCovariates(cvrt)



### First, illustrate with numerical phenotype
# Numerical, 3 value phenotype
phenotype = rep(1:3, each = 10)

# Test for association
output = testPhenotype(phenotype, data, cvrtqr)

# Show the results
print(output)

# Comparing with standard R code for the first variable
summary(lm( data[,1] ~ phenotype + cvrt ))



### First, illustrate with numerical phenotype
# Categorical, 3 group phenotype
phenotype = rep(c("Normal", "Sick", "Dead"), each = 10)

# Test for association
output = testPhenotype(phenotype, data, cvrtqr)

# Show the results
print(output)

# Comparing with standard R code for the first variable
anova(lm( data[,1] ~ cvrt + phenotype ))

andreyshabalin/ramwas documentation built on Sept. 27, 2021, 7:25 p.m.