Package index • attrition

Bounds on the average treatment effect

Worst-case bounds from a single round of data collection, and the double-sampling bounds that narrow them by chasing a random sample of the nonrespondents. All three accept a stratification variable for poststratification, which targets the same identification region and estimates it more precisely.

estimator_ev(): Extreme Value (Manski) Bounds
estimator_ds(): Extreme Value Bounds with Double Sampling
estimator_ds_sens(): Extreme Value Bounds with Double Sampling with Sensitivity

Sensitivity analysis

How far ignorability has to fail among the follow-up nonrespondents before a finding stops holding.

sensitivity_ds(): Sensitivity Analysis

Trimming bounds

Bounds on the effect among subjects who would report either way, under monotone selection in a stated direction (Lee 2009) or under random assignment alone (Imai 2008). The outcome need not be bounded, and either assumption pairs with either design.

estimator_trim(): Trimming Bounds

Working with results

Every estimator returns a named numeric vector. The tidy methods put the same quantities in a data frame with the column names DeclareDesign expects.

tidy(<attrition_bounds>): Tidy an attrition bounds object
tidy(<attrition_trim>): Tidy a trimming bounds object
tidy(<attrition_sensitivity>): Tidy a sensitivity analysis
summary(<attrition_bounds>): Summarize bounds
summary(<attrition_trim>): Summarize trimming bounds
print(<attrition_bounds>): Print bounds
print(<attrition_trim>): Print trimming bounds
print(<attrition_sensitivity>): Print a sensitivity analysis
reexports tidy: Objects exported from other packages

Data

levendusky_replication: Perceived polarization under double sampling

Package overview

attrition attrition-package: attrition: bounds for experiments with missing outcomes