jaguar_sim: Run null or power simulations
In cramanuj/JAGUAR: Joint Analysis of Genotype and Group-Specific Variability Using a Novel Score Test Approach to Map Expression Quantitative Trait Loci (eQTL)

Description Usage Arguments Details Value Author(s) References See Also Examples

Function to run power/null simulations by simulating one gene and one SNP at a time. The objective of these simulations is two pronged - 1) Check for the type I error control for the joint score test statistic, and 2) Compare two different null hypotheses where one's called a global null (bta=0 and PVEg =0) and other is local null (PVEg=0). Under the global null hypotheses, we fit a model where we assume that there is no main genotypic effect and group-specific variability in the data. Under the local null, we fit a model where we assume only the absence of group-specific variability. This is essentially a variance component score test.

1	jaguar_sim(nobs = 500, k = 5, tau = 1, eps = 1, PVEg = 0, bta = 0, maf = 0.10)

`nobs`	The number of observations in each group
`k`	The total number of groups
`tau`	Variance component of the subject-specific random effect
`eps`	Variance component of the residual error
`PVEg`	Proportion of variance explained by gamma
`bta`	Additive genotypic effect as a fixed-effect
`maf`	Minor allele frequency

This function currently implements only balanced designs with equal number of observations in each group. For each individual, we model the potential genetic association between a target SNP and the expression level a target gene (in multiple tissues) at a single locus using the following mixed effects model (i = individual; t = tissue) -

y_{i,t}=α_t + g_i β_i + b_t g_i + u_i + ε_{i,t}

where y_{i,t} is a t-dimensional vector of gene expression data for individual i, g_i is the scalar value of genotype in allele-dosage format, b_t is a t-dimensional tissue-specific random effect where b \sim N(0,γ), u_i is the scalar value representing individual-specific random effect where u \sim(0,τ).

A numeric vector consisting of two different p-values, "VCScoreTest" and "JointScoreTest" with the former indicating the p-value from the variance component score test and the latter indicating the p-value from the joint score test.

Chaitanya R. Acharya, Andrew S. Allen Maintainer: Chaitanya Acharya<c.acharya@duke.edu>

Chaitanya R. Acharya, Kouros Owzar, Janice M. McCarthy and Andrew S. Allen; Exploiting expression patterns across multiple tissues to map expression quantitative trait loci (Manuscript submitted)

Chaitanya R. Acharya and Andrew S. Allen; JAGUAR: An R-package to Implement Joint Analysis of Genotype and Group-Specific Variability Using a Novel Score Test to Map eQTL (Manuscript submitted)

jaguar_process,jaguar_slice,jaguar_gwa,jaguar_cis,jaguar_plotqtl

## An example to perform some null simulations
## NOTE: 10 sims are obviously not enough. Please try between 1000-10000.

nsim=10; alpha=0.05;
test = do.call("rbind",rlply(nsim,.progress="none",jaguar_sim(nobs=100,k=4)));
null.sim = apply(test,2,function(x) sum(x<=alpha)/nsim);