qgcomp.tobit.noboot: Quantile g-computation for left-censored outcomes

In qgcomp: Quantile G-Computation

Quantile g-computation for left-censored outcomes

Description

This function yields population average effect estimates for (possibly left censored) outcomes #'

Usage

qgcomp.tobit.noboot(
  f,
  data,
  expnms = NULL,
  q = 4,
  breaks = NULL,
  id = NULL,
  weights = NULL,
  cluster = NULL,
  alpha = 0.05,
  left = -Inf,
  right = Inf,
  ...
)

Arguments

f	an r formula representing the conditional model for the outcome, given all exposures and covariates. Interaction terms that include exposure variables should be represented via the AsIs function
data	data frame
expnms	character vector of exposures of interest
q	NULL or number of quantiles used to create quantile indicator variables representing the exposure variables. If NULL, then gcomp proceeds with un-transformed version of exposures in the input datasets (useful if data are already transformed, or for performing standard g-computation)
breaks	(optional) NULL, or a list of (equal length) numeric vectors that characterize the minimum value of each category for which to break up the variables named in expnms. This is an alternative to using 'q' to define cutpoints.
id	Not used
weights	"case weights" - passed to the "weight" argument of AER::tobit. Note this must be a standalone vector so that it might accept, for example, an argument like data$weights but will not accept "weights" if there is a variable "weights" in the dataset representing the weights. tobit
cluster	Passed to AER::tobit (Optional variable that identifies groups of subjects, used in computing the robust variance. Like model variables, this is searched for in the dataset pointed to by the data argument). Note: this will result in robust standard errors being returned unless robust=FALSE is used in the function call.
alpha	alpha level for confidence limit calculation
left	Passed to AER::tobit (From tobit docs: left limit for the censored dependent variable y. If set to -Inf, y is assumed not to be left-censored.)
right	Passed to AER::tobit (From tobit docs: right limit for the censored dependent variable y. If set to Inf, the default, y is assumed not to be right-censored.)
...	arguments to AER::tobit (e.g. dist), and consequently to survival::survreg

Value

a qgcompfit object, which contains information about the effect measure of interest (psi) and associated variance (var.psi), as well as information on the model fit (fit) and information on the weights/standardized coefficients in the positive (pos.weights) and negative (neg.weights) directions.

Examples

# First, a demonstration that the tobit model will return identical results 
# (with variation due to numerical algorithms)
# in the case of no censored values
set.seed(50)
# linear model
dat <- data.frame(y=runif(50,-1,1), x1=runif(50), x2=runif(50), z=runif(50))
qgcomp.glm.noboot(f=y ~ z + x1 + x2, expnms = c('x1', 'x2'), data=dat, q=2, family=gaussian())
qgcomp.tobit.noboot(f=y ~ z + x1 + x2, expnms = c('x1', 'x2'), data=dat, q=2)
# not intercept model
qgcomp.glm.noboot(f=y ~-1+ z + x1 + x2, expnms = c('x1', 'x2'), data=dat, 
    q=2, family=gaussian())
qgcomp.tobit.noboot(f=y ~-1+ z + x1 + x2, expnms = c('x1', 'x2'), data=dat, 
    q=2, dist="gaussian", left = -Inf, right = Inf)
# Next, a demonstration that the tobit model address censoring
uncens_dat = qgcomp:::.dgm_quantized(N=1000, b0=0, coef=c(.33,.33,.33,.0))
uncens_dat$ycens = ifelse(uncens_dat$y>0, uncens_dat$y, 0) # censor at zero
uncens_dat$ycens1 = ifelse(uncens_dat$y>0, uncens_dat$y, 1) # censor at higher value
# q set to NULL because the exposures are already quantized
# again, first check the uncensored outcome for agreement
qgcomp.glm.noboot(f= y ~ x1+x2+x3+x4, expnms = c('x1', 'x2', 'x3', 'x4'), 
    data=uncens_dat, q=NULL, family=gaussian())
qgcomp.tobit.noboot(f= y ~x1+x2+x3+x4, expnms = c('x1', 'x2', 'x3', 'x4'), 
    data=uncens_dat, q=NULL, dist="gaussian", left = -Inf, right = Inf)
# Next, after censoring
ft_std <- qgcomp.glm.noboot(f= ycens ~ x1+x2+x3+x4, 
    expnms = c('x1', 'x2', 'x3', 'x4'), data=uncens_dat, q=NULL, 
    family=gaussian())
ft_tobit <- qgcomp.tobit.noboot(f= ycens ~x1+x2+x3+x4, 
    expnms = c('x1', 'x2', 'x3', 'x4'), data=uncens_dat, q=NULL, 
    dist="gaussian", left = 0, right = Inf)
# note the tobit regression will be closer to the estimates given when 
# fitting with the uncensored outcome (which will typically not be available)
summary(ft_std) 
summary(ft_tobit)
ft_std1 <- qgcomp.glm.noboot(f= ycens1 ~ x1+x2+x3+x4, 
    expnms = c('x1', 'x2', 'x3', 'x4'), data=uncens_dat, q=NULL, 
    family=gaussian())
ft_tobit1 <- qgcomp.tobit.noboot(f= ycens1 ~x1+x2+x3+x4, 
    expnms = c('x1', 'x2', 'x3', 'x4'), data=uncens_dat, q=NULL, 
    dist="gaussian", left = 1, right = Inf)
# the bias from standard methods is more extreme at higher censoring levels
summary(ft_std1) 
summary(ft_tobit1)

qgcomp documentation built on April 12, 2025, 2:28 a.m.