Skip to contents

Estimate bias due to sample overlap

Source: R/estimate_overlap_bias.R
estimate_overlap_bias.Rd

Estimate bias due to sample overlap

Usage

estimate_overlap_bias(
  samplesize_exposure,
  samplesize_outcome,
  n_variants,
  rsq_exposure,
  exp_f = NULL,
  lci_95 = FALSE,
  case_prop = 0,
  ols_bias,
  overlap_prop,
  var_x = 1,
  var_y = 1
)

Arguments

samplesize_exposure

(numeric) Sample size of population used to define genetic instrument for the exposure of interest

samplesize_outcome

(numeric) Sample size of population used for the outcome of interest

n_variants

(numeric) Number of genetic variants included in genetic instrument for the exposure of interest

rsq_exposure

(numeric) \(R^2\) value (coefficient of determination) of genetic instrument for the exposure of interest; used to estimate F-statistic

exp_f

(numeric; optional) F-statistic for the genetic instrument (if provided, this value will be used, rather than an estimate based on the \(R^2\))

lci_95

(logical; default = FALSE) If TRUE, the function will return estimates of bias and type 1 error based on the lower limit of the one-sided 95% confidence interval of the F-statistic, which may represent a more conservative/less optimistic estimate of bias

case_prop

(numeric; optional) Proportion of cases (eg. cases/total samplesize) if outcome is binary

ols_bias

(numeric) Observational (biased) effect estimate (if known); otherwise, provide a hypothetical value

overlap_prop

(numeric; range = 0 to 1) Proportion of overlapping samples between exposure and outcome studies (if known); otherwise, provide a hypothetical value

var_x

(numeric) Variance in the exposure; default is 1 when the exposure is reported in standard deviation units

var_y

(numeric) Variance in the exposure; default is 1 when the exposure is reported in standard deviation units

Value

A tibble containing columns for the bias and type1_error

Examples

# Binary outcome
estimate_overlap_bias(
  samplesize_exposure = 361194,
  samplesize_outcome = 1125328,
  case_prop = 0.035,
  rsq_exposure = 0.068,
  n_variants = 196,
  ols_bias = 0.2,
  overlap_prop = 0.3
)
#> # A tibble: 1 × 2
#>       bias type1_error
#>      <dbl>       <dbl>
#> 1 0.000446      0.0501

# Continuous outcome
estimate_overlap_bias(
  samplesize_exposure = 361194,
  samplesize_outcome = 1125328,
  rsq_exposure = 0.068,
  n_variants = 196,
  ols_bias = 0.2,
  overlap_prop = 0.3
)
#> # A tibble: 1 × 2
#>       bias type1_error
#>      <dbl>       <dbl>
#> 1 0.000446      0.0517