interval_hypothesis: Hypothesis tests for identification regions

In brentonk/coefbounds: Nonparametric Bounds for Regression with Interval-Censored Outcomes

Description

Given an interval B = [b_L, b_U], tests against the null hypothesis that the identification region for the specified coefficient is a superset of B (type = "subset", the default) or equals B (type = "equal").

Usage

interval_hypothesis(fit, term, interval, type = c("subset", "equal"))

Arguments

fit	Fitted model of class "coefbounds". Must have been run with boot > 0 and contain a non-null dist element.
term	Name of the coefficient whose identification region is being tested.
interval	Interval specified in the null hypothesis. May be a singleton; e.g., interval = c(0, 0) to test that the identification region contains 0.
type	Whether the null hypothesis should be that interval is a "subset" of the population identification region (the default) or is "equal" to it.

Details

The test with type = "subset" corresponds to confint.coefbounds with type = "DU". The test with type = "equal" corresponds to confint.coefbounds with type = "CC".

Implements largely the same functionality as CI1D in Beresteanu et al.'s (2010) Stata program.

Value

A list of class "interval_hypothesis" containing:

test_stat

Test statistic.

p

Bootstrap p-value.

n_boot

Number of bootstrap iterations.

clus

Whether a cluster bootstrap was used.

term

Name of coefficient tested.

interval

Interval specified.

estimate

Sample estimate of coefficient bounds.

type

Type of test performed.

Author(s)

Brenton Kenkel

References

Arie Beresteanu, Francesca Molinari and Darcy Steeg Morris. 2010. "Asymptotics for Partially Identified Models in Stata." https://molinari.economics.cornell.edu/programs.html

Examples

## Simulate data
set.seed(18)
x1 <- rnorm(50)
x2 <- rnorm(50)
y <- 1 - x1 + x2 + rnorm(50)
yl <- floor(y)
yu <- ceiling(y)

## Fit model
fit <- coefbounds(yl + yu ~ x1 + x2, boot = 100)

## Test hypothesis that the identification region for the coefficient on x1
## contains -0.25
interval_hypothesis(fit,
                    term = "x1",
                    interval = c(-0.25, -0.25),
                    type = "subset")

## Test hypothesis that the identification region for the coefficient on x2
## equals [0, 1.5]
interval_hypothesis(fit,
                    term = "x2",
                    interval = c(0, 1.5),
                    type = "equal")

brentonk/coefbounds documentation built on May 13, 2019, 5:09 a.m.