dpit_tobit: Residuals for a tobit model

In assessor: Assessment Tools for Regression Models with Discrete and Semicontinuous Outcomes

Residuals for a tobit model

Description

Computes DPIT residuals for tobit regression models using the observed responses (y) and their corresponding fitted distributional parameters (mu, sd).

Usage

dpit_tobit(y, mu, sd)

Arguments

y	An observed outcome vector.
mu	A vector of fitted mean values of latent variables.
sd	A standard deviation of latent variables.

Details

For formulation details on semicontinuous outcomes, see dpit.

Value

The dpit object containing DPIT residuals.

Examples

## Tobit regression model
library(VGAM)
n <- 500
beta13 <- 1
beta14 <- -3
beta15 <- 3

set.seed(1234)
x11 <- runif(n)
x12 <- runif(n)
lambda1 <- beta13 + beta14 * x11 + beta15 * x12
sd0 <- 0.3
yun <- rnorm(n, mean = lambda1, sd = sd0)
y <- ifelse(yun >= 0, yun, 0)

# Using VGAM package
# True model
fit1 <- vglm(formula = y ~ x11 + x12,
             tobit(Upper = Inf, Lower = 0, lmu = "identitylink"))
# Missing covariate
fit1miss <- vglm(formula = y ~ x11,
                 tobit(Upper = Inf, Lower = 0, lmu = "identitylink"))

dpit.tobit1 <- dpit_tobit(y = y, mu = VGAM::fitted(fit1), sd = sd0)
resid.tobit1 <- residuals(dpit.tobit1)
plot(dpit.tobit1)
dpit.tobit2 <- dpit_tobit(y = y, mu = VGAM::fitted(fit1miss), sd = sd0)
resid.tobit2 <- residuals(dpit.tobit2)
plot(dpit.tobit2)

# Using AER package
library(AER)
# True model
fit2 <- tobit(y ~ x11 + x12, left = 0, right = Inf, dist = "gaussian")
# Missing covariate
fit2miss <- tobit(y ~ x11, left = 0, right = Inf, dist = "gaussian")

dpit.aer1 <- dpit_tobit(y = y, mu = fitted(fit2), sd = sd0)
resid.aer1 <- residuals(dpit.aer1)
plot(dpit.aer1)
dpit.aer2 <- dpit_tobit(y = y, mu = fitted(fit2miss), sd = sd0)
resid.aer2 <- residuals(dpit.aer2)
plot(dpit.aer2)

assessor documentation built on Aug. 22, 2026, 9:06 a.m.