konfound: Perform sensitivity analysis on fitted models

In konfound: Quantify the Robustness of Causal Inferences

Description

For fitted models, this command calculates (1) how much bias there must be in an estimate to invalidate/sustain an inference; (2) the impact of an omitted variable necessary to invalidate/sustain an inference for a regression coefficient. Currently works for: models created with lm() (linear models).

Usage

konfound(
  model_object,
  tested_variable,
  alpha = 0.05,
  tails = 2,
  index = "RIR",
  to_return = "print",
  test_all = FALSE,
  two_by_two = FALSE,
  n_treat = NULL,
  switch_trm = TRUE,
  replace = "control"
)

Arguments

model_object	output from a model (currently works for: lm)
tested_variable	Variable associated with the unstandardized beta coefficient to be tested
alpha	probability of rejecting the null hypothesis (defaults to 0.05)
tails	integer whether hypothesis testing is one-tailed (1) or two-tailed (2; defaults to 2)
index	whether output is RIR or IT (impact threshold); defaults to "RIR"
to_return	whether to return a data.frame (by specifying this argument to equal "raw_output" for use in other analyses) or a plot ("plot"); default is to print ("print") the output to the console; can specify a vector of output to return
test_all	whether to carry out the sensitivity test for all of the coefficients (defaults to FALSE)
two_by_two	whether or not the tested variable is a dichotomous variable in a GLM; if so, the 2X2 table approach is used; only works for single variables at present (so test_all = TRUE will return an error)
n_treat	the number of cases associated with the treatment condition; applicable only when model_type = "logistic"
switch_trm	whether to switch the treatment and control cases; defaults to FALSE; applicable only when model_type = "logistic"
replace	whether using entire sample or the control group to calculate the base rate; default is the entire sample

Value

prints the bias and the number of cases that would have to be replaced with cases for which there is no effect to invalidate the inference

Examples

# using lm() for linear models
m1 <- lm(mpg ~ wt + hp, data = mtcars)
konfound(m1, wt)
konfound(m1, wt, test_all = TRUE)
konfound(m1, wt, to_return = "table")

# using glm() for non-linear models
if (requireNamespace("forcats")) {
  d <- forcats::gss_cat

  d$married <- ifelse(d$marital == "Married", 1, 0)

  m2 <- glm(married ~ age, data = d, family = binomial(link = "logit"))
  konfound(m2, age)
}

# using lme4 for mixed effects (or multi-level) models
if (requireNamespace("lme4")) {
  library(lme4)
  m3 <- fm1 <- lme4::lmer(Reaction ~ Days + (1 | Subject), sleepstudy)
  konfound(m3, Days)
}

m4 <- glm(outcome ~ condition, data = binary_dummy_data, family = binomial(link = "logit"))
konfound(m4, condition, two_by_two = TRUE, n_treat = 55)

konfound documentation built on June 1, 2021, 9:08 a.m.