RVS_asy: RVS using asymptotic distribution for score test statistic

In Struglab/RVS: Robust Variance Score (RVS) test for association of variants from next generation sequence data with binary traits using external controls

Description

This functions actually includes two association tests which differ only at the estimation of the variance of score statistic by option 'method'. It uses regular formula for Var_{case}(E(G_{ij}|D_{ij})) when method = 'Regular' (also called likelihood method in the paper), Var_{case}(G_{ij}) is used for Var_{case}(E(G_{ij}|D_{ij})) when method = 'RVS'. It uses the population frequency P to estimate variance for case if method=RVS, otherwise use the expected genotype probability for case directly.

Usage

RVS_asy(Y, X, P, RVS = "TRUE")

Arguments

Y	- a vector with the phenotype for n individuals (=ncase+ncont), first ncase (y=1) then ncont (y=0) (integar)
X	- a vector of genotype for a snp, first ncase and then ncont
P	(vector with length 3 (double)) - estimated genotype frequency for a variant P(G=0), P(G=1) and P(G=2)
RVS,	logic variable to indicate the estimation of variance of the score statistic.

Details

Note, both function scaled the variance in equation (1) in Appendix A by dividing N_case*Ncontrol/N also note the test statistics in anova gives us Rao=s^2/var(s^2), where RVS=s^2/robust var(s^2) so RVS=Rao *var(s^2)/robust var(s^2), s=sum(y_j-mean(y))E(G_ij|D_ij), therefore var(s)//var(E(Gij|Dij)), var(X) in code.

Value

p-values for the snp

References

http://www.ncbi.nlm.nih.gov/pubmed/22570057

http://www.ncbi.nlm.nih.gov/pubmed/24733292

Struglab/RVS documentation built on May 9, 2019, 3:11 p.m.