Simulate MR Data with Known Causal Effect

Generates a complete dataset with SNP genotypes, an exposure, confounders, and a binary outcome where the true causal effect of the exposure on the outcome is known and recoverable. Useful for testing and vignette examples without requiring a database connection.

Usage

simulateMRData(
  n = 5000,
  nSnps = 10,
  trueEffect = 0.5,
  confoundingStrength = 0.3,
  snpEffectRange = c(0.1, 0.5),
  seed = 42
)

Arguments

n

Number of individuals.

nSnps

Number of SNPs (instruments).

trueEffect

True causal effect of exposure on outcome (log-OR scale).

confoundingStrength

Strength of confounding (coefficient of confounder on both exposure and outcome).

snpEffectRange

Range of SNP-exposure effect sizes.

seed

Random seed for reproducibility.

Value

A list with elements:

data

Data frame with person_id, outcome (0/1), snp_<sanitized rsID> columns, confounder_1, confounder_2, exposure.

instrumentTable

Data frame mimicking getMRInstruments() output.

trueEffect

The true causal effect used in simulation.

Examples

simData <- simulateMRData(n = 1000, nSnps = 5, trueEffect = 0.3)
head(simData$data)
#>   person_id outcome snp_rs1 snp_rs2 snp_rs3 snp_rs4 snp_rs5 confounder_1
#> 1         1       1       1       1       0       0       1    1.3709584
#> 2         2       0       1       0       1       0       0   -0.5646982
#> 3         3       1       1       0       2       1       0    0.3631284
#> 4         4       1       0       0       1       1       0    0.6328626
#> 5         5       1       0       2       0       0       0    0.4042683
#> 6         6       1       0       1       0       0       0   -0.1061245
#>   confounder_2   exposure
#> 1            1  3.3068319
#> 2            0 -0.4567887
#> 3            1  1.8067824
#> 4            1  0.7664246
#> 5            1  3.0807866
#> 6            1 -0.1501653