snpStats (version 1.22.0)

single.snp.tests: 1-df and 2-df tests for genetic associations with SNPs (or imputed SNPs)

Description

This function carries out tests for association between phenotype and a series of single nucleotide polymorphisms (SNPs), within strata defined by a possibly confounding factor. SNPs are considered one at a time and both 1-df and 2-df tests are calculated. For a binary phenotype, the 1-df test is the Cochran-Armitage test (or, when stratified, the Mantel-extension test). The function will also calculate the same tests for SNPs imputed by regression analysis.

Usage

single.snp.tests(phenotype, stratum, data = sys.parent(), snp.data, rules=NULL, subset, snp.subset, uncertain = FALSE, score=FALSE)

Arguments

phenotype

A vector containing the values of the phenotype

stratum

Optionally, a factor defining strata for the analysis

data

A dataframe containing the phenotype and stratum data. The row names of this are linked with the row names of the snps argument to establish correspondence of phenotype and genotype data. If this argument is not supplied, phenotype and stratum are evaluated in the calling environment and should be in the same order as rows of snps

snp.data

An object of class "SnpMatrix" containing the SNP genotypes to be tested

rules

An object of class "ImputationRules". If supplied, the rules coded in this object are used, together with snp.data, to calculate tests for imputed SNPs

subset

A vector or expression describing the subset of subjects to be used in the analysis. This is evaluated in the same environment as the phenotype and stratum arguments

snp.subset

A vector describing the subset of SNPs to be considered. Default action is to test all SNPs in snp.data or, in imputation mode, as specified by rules

uncertain

If TRUE, uncertain genotypes are handled by replacing score contributions by their posterior expectations. Otherwise they are treated as missing. Setting this option authomatically invokes use of robust variance estimates

score

If TRUE, the output object will contain, for each SNP, the score vector and its variance-covariance matrix

Value

An object of class "SingleSnpTests". If score is set to TRUE, the output object will be of the extended class "SingleSnpTestsScore" containing additional slots holding the score statistics and their variances (and covariances). This allows meta-analysis using the pool function.

Details

Formally, the test statistics are score tests for generalized linear models with canonical link. That is, they are inner products between genotype indicators and the deviations of phenotypes from their stratum means. Variances (and covariances) are those of the permutation distribution obtained by randomly permuting phenotype within stratum.

When the function is used to calculate tests for imputed SNPs, the test is still a score test. The score statistics are calculated from the expected value, given observed SNPs, of the score statistic if the SNP to be tested were itself observed. The subset argument can either be a logical vector of length equal to the length of the vector of phenotypes, an integer vector specifying positions in the data frame, or a character vector containing names of the selected rows in the data frame. Similarly, the snp.subset argument can be a logical, integer, or character vector.

References

Chapman J.M., Cooper J.D., Todd J.A. and Clayton D.G. (2003) Human Heredity, 56:18-31. Clayton (2008) Testing for association on the X chromosome Biostatistics, 9:593-600.)

See Also

snp.lhs.tests, snp.rhs.tests, impute.snps, ImputationRules-class, pool, SingleSnpTests-class, SingleSnpTestsScore-class

Examples

data(testdata)
results <- single.snp.tests(cc, stratum=region, data=subject.data,
   snp.data=Autosomes, snp.subset=1:10)
print(summary(results))

# writing to an (anonymous and temporary) csv file
csvfile <- tempfile()
write.csv(file=csvfile, as(results, 'data.frame'))
unlink(csvfile)
# QQ plot 
qq.chisq(chi.squared(results, 1), 1)
qq.chisq(chi.squared(results, 2), 2)