robust_outcome: Implements the robust modified outcome approach

In epiGWAS: Robust Methods for Epistasis Detection

Description

A key feature of robust_outcome is its resilience to the misspecification of propensity scores, which is a major limitation of classical modified outcome approaches. Except for shifted_outcome, all of the modified outcome approaches belong to a parameterized class of unbiased estimators for the risk difference term E[Y|A=1,X]-E[Y|A=0,X]. Within that class, robust modified outcome is the approach with the least large-sample variance. For more details about this approach, see Lunceford and Davidian (2004)

Usage

robust_outcome(A, X, Y, propensity, parallel = FALSE, ...)

Arguments

A	target variant
X	rest of the genotype
Y	phenotype
propensity	propensity scores
parallel	whether to perform support estimation in a parallelized fashion
...	additional arguments to be passed to stabilityGLM or stabilityBIG

Value

a vector containing the area under the stability selection path for each variable in X

References

Lunceford, J. K., & Davidian, M. (2004). Stratification and weighting via the propensity score in estimation of causal treatment effects: A comparative study. Statistics in Medicine, 23(19), 2937–2960.

Examples

n <- 30
p <- 10
X <- matrix((runif(n * p) < 0.4) + (runif(n * p) < 0.4),
             ncol = p, nrow = n) # SNP matrix
A <- rbinom(n, 1, 0.3)
propensity <- runif(n, min = 0.4, max = 0.8)
Y <- runif(n) < 0.4
robust_scores <- robust_outcome(A, X, Y, propensity,
                               lambda_min_ratio = 0.01 , n_subsample = 1)

epiGWAS documentation built on Sept. 8, 2019, 5:02 p.m.