selection_bound: Sensitivity analysis for selection bias

In matt-tudball/selectioninterval: Selection Bias Sensitivity Analysis

Sensitivity analysis for selection bias

Description

Given a set of sensitivity parameters and constraints, computes an upper and lower bound for an inverse probability weighted regression estimate

Usage

selection_bound(
  y,
  x,
  w,
  z = NULL,
  L0l,
  L0u,
  L1,
  cons = NULL,
  theta = NULL,
  alpha = 0.05,
  opts = NULL
)

Arguments

y	Outcome (vector)
x	Explanatory variables (matrix)
w	Selection variables (matrix)
z	Optional instrumental variables (matrix)
L0l	Lower bound for the probability of sample selection for the average observation
L0u	Upper bound for the probability of sample selection for the average observation
L1	Odds ratio of sample selection for a one unit (binary) or one standard deviation (continuous) increase in a variable.
cons	List of constraints to be applied: RESP is a response rate constraint; COVMEAN is a covariate mean constraint; DIREC is a directionality constraint.
theta	Optional starting parameter for the global optimiser.
alpha	Significance level for confidence interval.
opts	Optional list of options for the global optimiser.

Value

Named list of objects: theta_min (theta_max) is the parameters for the lower (upper) bound; interval is a vector containing the lower and upper bounds; ci is a vector containing the alpha-level confidence interval.

matt-tudball/selectioninterval documentation built on Sept. 4, 2022, 4:08 p.m.