epiGWAS (version 1.0.2)

OWL: Implements the outcome weighted learning approach

Description

To recover the synergistic interactions between the target A and the rest of the genotype X, OWL formulates a weighted binary classification problem. The outcome is the mapping of A to {0,1}. The covariates are X. The propensity scores and the phenotypes are combined in the sample weights \(Y/\pi(A| X)\). For binary phenotypes, OWL is a case-only approach. The approach also accommodates nonnegative continuous phenotypes.

Usage

OWL(A, X, Y, propensity, ...)

Arguments

A

target variant. If not binary, the variable A must be encoded as either (0, 1) or (0, 1, 2).

X

rest of the genotype

Y

phenotype (binary or continuous)

propensity

propensity scores (a vector or a two-column matrix)

...

additional arguments to stabilityGLM

Value

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

Details

For continuous phenotypes, if the outcome Y is not nonnegative, it is translated to make it nonnegative.

References

Zhao, Y., Zeng, D., Rush, A. J., & Kosorok, M. R. (2012). Estimating Individualized Treatment Rules Using Outcome Weighted Learning. Journal of the American Statistical Association, 107(499), 1106<U+2013>1118.

Examples

# NOT RUN {
n <- 30
p <- 10
X <- matrix((runif(n * p) < 0.5) + (runif(n * p) < 0.5), ncol = p, nrow = n)
A <- (runif(n, min = 0, max = 1) < 0.3)
propensity <- runif(n, min = 0.4, max = 0.8)
Y <- runif(n, min = 0, max = 1) < 1/ (1 + exp(-X[, c(1, 7)] %*% rnorm(2)))
OWL(A, X, Y, propensity, short = FALSE, n_lambda = 50, n_subsample = 1)

# }