Skip to contents

Simulate target / source data under the four-group model

Source: R/simulate.R
heart_simulate.Rd

Generates a toy two-population data set for examples and tests. Variants are independent; each variant is assigned to one of four configurations with probabilities prop: "00" (null in both populations), "01" (source-specific signal: target effect zero, source effect non-zero), "10" (target-specific signal) and "11" (shared signal, with source effect equal to the target effect times a random factor around 1). Allele frequencies and covariate distributions differ between populations; the genetic contribution is centred within each population so that the pooled analysis is free of population stratification (in real data, include genetic principal components in X). A surrogate S that is informative of Y is also returned so that the doubly robust estimator can be illustrated.

Usage

heart_simulate(
  n_tar = 2000,
  n_src = 20000,
  J = 2000,
  prop = c(`00` = 0.9, `01` = 0.04, `10` = 0.01, `11` = 0.05),
  h2_snp = 0.004,
  seed = NULL
)

Arguments

n_tar, n_src

Target and source sample sizes.

J

Number of variants.

prop

Probabilities of the configurations c("00", "01", "10", "11").

h2_snp

Variance explained by each causal variant in its population.

seed

Optional random seed.

Value

A list with Y, D (1 = target), X (two covariates), G (n x J genotype matrix), S (surrogate) and truth, a data.frame with the configuration and the true target and source effects of each variant.

Examples

sim <- heart_simulate(n_tar = 500, n_src = 2000, J = 50, seed = 1)
table(sim$truth$config)
#> 
#> 00 01 10 11 
#> 43  3  1  3