tobit1: Truncated response model

In tobit1: One Equation Tobit Model

Truncated response model

Description

Estimation of models for which the response is truncated, either on censored or truncated samples using linear models, maximum likelihood or two-steps estimators

Usage

tobit1(
  formula,
  data,
  subset = NULL,
  weights = NULL,
  start = NULL,
  left = 0,
  right = Inf,
  scedas = c("exp", "pnorm"),
  sample = c("censored", "truncated"),
  method = c("ml", "lm", "2steps", "trimmed", "nls"),
  trace = FALSE
)

## S3 method for class 'tobit1'
nobs(object, ...)

## S3 method for class 'tobit1'
vcov(object, ...)

## S3 method for class 'tobit1'
logLik(object, ...)

## S3 method for class 'tobit1'
summary(object, ...)

## S3 method for class 'tobit1'
print(x, digits = max(3L, getOption("digits") - 3L), ...)

## S3 method for class 'summary.tobit1'
print(
  x,
  digits = max(3, getOption("digits") - 2),
  width = getOption("width"),
  ...
)

Arguments

formula	a symbolic description of the model; if two right hand sides are provided, the second one is used to parametrize the conditional variance,
data	a data frame,
subset	a subset,
weights	an optional vector of weights (currently only supported by ml method
start	an optional vector of starting values
left, right	left and right limits of the dependent variable. The default is respectively 0 and +Inf which corresponds to the most classic (left-zero truncated) tobit model,
scedas	the functional form used to specify the conditional variance, which is of the form: s_n = s_o f(Z'g), where Z are the covariates indicated in the second part of the formula and z_o and g a set of parameters to estimate. Currently, f can either be set to "exp" or "pnorm",
sample	either "censored" (the default) to estimate the censored (tobit) regression model or "truncated" to estimated the truncated regression model,
method	one of "ml" for maximum likelihood, "lm" for (biased) least squares estimators and "2steps" for two-steps consistent estimators, "trimmed" for symetrically censored estimator,
trace	a boolean (the default if FALSE) if TRUE some information about the optimization process is printed,
...	further arguments.
x, object	an object of class tobit1 or summary.tobit1,
digits, width	see base::print,

Value

An object of class c('tobit1', 'lm'), which is a list containg the following components:

• coefficients: a named vector of coefficients,

• linear.predictor: the linear fit,

• fitted.values: the fitted values,

• residuals: the residuals,

• df.residual: the residual degrees of freedom,

• hessian: the hessian of the log-likelihood function at the optimum,

• vcov: an estimator of the covariance matrix of the coefficients,

• gradObs: a N x K matrix containing the individual contributions to the gradient,

• logLik: the value of the log-likelihood at the optimum,

• model: the model frame,

• terms: the terms object used,

• call: the matched call

• xlevels: a record of the levels of the factors used in fitting

• na.action: intormation returned by model.frame on the special handling of NA's.

Author(s)

Yves Croissant

Examples

# tobit model estimated by maximum likelihood
tobit1(fees ~ expense, feesadm)
# same using two-steps estimator
tobit1(fees ~ expense, feesadm, method = "2steps")
# same model fitted on the truncated sample
tobit1(fees ~ expense, feesadm, sample = "truncated")

tobit1 documentation built on March 18, 2022, 7:31 p.m.