R: Causal Effect Based on Semiparametric Least Squares Models

causalSLSE {causalSLSE}

R Documentation

Causal Effect Based on Semiparametric Least Squares Models

Description

This is the main method to estimate the causal effects using the semiparametric least squares method. It returns an object of class cslse and is registered for objects of class cslseModel and cslseFit.

Usage

## S3 method for class 'cslseModel'
causalSLSE(object,
                               selType=c("SLSE","BLSE","FLSE"),
                               selCrit = c("AIC", "BIC", "PVT"),
                               selVcov = c("HC0", "Classical", "HC1", "HC2", "HC3"),    
                               causal = c("ALL","ACT","ACE","ACN"),
                               pvalT = function(p) 1/log(p),
                               vcov.=vcovHC, reSelect=FALSE, ...)
## S3 method for class 'cslseFit'
causalSLSE(object, causal = c("ALL","ACT","ACE","ACN"),
                             vcov.=vcovHC, ...)
## S3 method for class 'formula'
causalSLSE(object, data, nbasis=function(n) n^0.3,
                             knots, 
                             selType=c("SLSE","BLSE","FLSE"),
                             selCrit = c("AIC", "BIC", "PVT"),
                             selVcov = c("HC0", "Classical", "HC1", "HC2", "HC3"),        
                             causal = c("ALL","ACT","ACE","ACN"),
                             pvalT = function(p) 1/log(p),
                             vcov.=vcovHC, reSelect=FALSE, ...)

Arguments

`object`	An object of class `cslseModel` created by the `cslseModel` function, `cslseFit` created by `estSLSE` or `formula`.
`data`	A `data.frame` with all variables included in `form`.
`nbasis`	A function to determined the number of basis functions. It has to be a function of one argument, the sample size.
`knots`	A list of knots for the treated and nontreated groups. The list must be named using the group names. Each element of the list is also a list of length equal to the number of confounders. The choice for each confounders is NULL for no knots or numeric for specific values. If missing, the knots are automatically generated.
`selType`	The method for selecting the knots. By default (SLSE), all knots from the model are used.
`selCrit`	The criterion to select the knots.
`causal`	What causal effect should we compute.
`pvalT`	A function to determine the p-value threshold for the significance of the coefficients. It has to be a function of one parameter, which is the average number of knots in the model. This value may differ across treatment group.
`selVcov`	The type of least squares covariance matrix used to compute the p-values needed for the selection.
`vcov.`	An alternative function to compute the covariance matrix of the least squares estimators. The default is `vcovHC`. This function is used to compute the standard errors of the causal effect estimators and the SLSE coefficients.
`reSelect`	By default, the stored selections are used. If `reSelect` is set to `TRUE`, the selection is re-computed.
`...`	Additional arguments to pass to `vcov.`

Value

It returns an object of class cslse, which inherits from the class cslseFit. It is a list with the following elements:

`treated`, `nontreated`	They are objects of class `slseFit` obtained by `estSLSE`.
`ACE`, `ACT`, `ACN`	Estimates of the average causal effect, the causal effect on the treated and the causal effect on the nontreated. Each of them is a vector of two elements: the estimate and its estimated standard error. All three are included only if the argument `causal` is set to `"ALL"`.

Also, the object contains the following additional attributes:

`treatedVar`	The name of the variable in the dataset that represents the treatment indicator.
`groupInd`	A named vector with the value of the treatment indicator corresponding do each treatment group.

Examples

data(simDat3)

## A causal SLSE model with the outcome Y
## the treatment indicator Z and the confounders X1, X2 and X1:X2

mod1 <- cslseModel(Y ~ Z | ~ X1 * X2, data = simDat3)

## The causal effects are estimated using the backward method and AIC criterion
## The HC1 type is used for the least squares covariance matrix

fit1 <- causalSLSE(mod1, selType = "BLSE", type = "HC1")

## This is the same for formula objects

fit2 <- causalSLSE(Y ~ Z | ~ X1 * X2, data = simDat3, selType = "BLSE", type = "HC1")