confint.coefbounds: Confidence regions for coefficient bounds

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

Description

Bootstrap estimates of confidence regions for the bounds on model coefficients.

Usage

## S3 method for class 'coefbounds'
confint(object, parm = NULL, level = 0.95,
  type = c("DU", "CC"), ...)

Arguments

object	Fitted model of class "coefbounds". Must have been run with boot > 0 and contain a non-null dist element.
parm	Names or indices of the coefficients for which to estimate confidence sets. If not specified, uses all coefficients.
level	Confidence level.
type	Whether to return the DU confidence set or the CC confidence collections. See Details for explanation.
...	Not used.

Details

Two types of confidence region are available, both defined in Beresteanu and Molinari (2008, Section 2). The default is "DU", the confidence region based on the directed Hausdorff distance. We would fail to reject the null hypothesis that B is a subset of the population identification region for the coefficient β_j if and only if B \subseteq DU(β_j). The other option is "CC", the confidence collection based on the undirected Hausdorff distance. We would fail to reject the null hypothesis that [b_L, b_U] equals the population identification region for β_j if and only if b_L \in CC_L(β_j) and b_U \in CC_U(β_j).

The "DU" confidence region corresponds to interval_hypothesis with type = "subset". The "CC" confidence region corresponds to interval_hypothesis with type = "equal".

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

Value

A matrix containing the range of the confidence region for each term requested. When type== "CC", there are two entries per term (range of the confidence region for the lower and upper bound).

Author(s)

Brenton Kenkel

References

Arie Beresteanu and Francesca Molinari. 2008. "Asymptotic Properties for a Class of Partially Identified Models." Econometrica 76 (4): 763–814.

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)

## Calculate DU confidence region for coefficient on x1
confint(fit, parm = "x1")

## Calculate CC confidence collections for all coefficients
confint(fit, type = "CC")

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