Skip to contents

E-value for a bias-scale effect estimate (native)

Source: R/stats_primitives_native.R
morie_evalue.Rd

The minimum strength of unmeasured confounding (on the risk-ratio scale) that could explain away an observed association, plus the E-value for the CI bound closest to the null. Handles RR, OR, HR (rare / common outcome), and OLS / standardized-mean-difference estimands via the published VanderWeele-Ding conversions. Reproduces EValue::evalue / EValue::evalues.*.

Usage

morie_evalue(
  est,
  type = "RR",
  lo = NULL,
  hi = NULL,
  rare = TRUE,
  sd = NULL,
  true = NULL
)

Arguments

est

Point estimate on the type scale.

type

One of "RR", "OR", "HR", "OLS", "MD".

lo, hi

Optional CI bounds (same scale as est).

rare

For OR/HR, whether the outcome is rare (default TRUE).

sd

For OLS, the outcome standard deviation.

true

Null value (default 1 for ratio scales, 0 for MD/OLS).

Value

A list with point (E-value for the estimate) and ci (E-value for the near-null CI bound, or NA).

References

Ding, P., & VanderWeele, T. J. (2016). Sensitivity analysis without assumptions. Epidemiology, 27(3), 368-377.

Examples

morie_evalue(3.9, "RR", lo = 2.4)
#> $point
#> [1] 7.263034
#> 
#> $ci
#> [1] 4.23303
#>