predict.DTRreg {DTRreg}

Optimal Outcome Prediction for DTRs

Description

Predicted outcome assuming optimal treatment (according to analysis via G-estimation or dWOLS) was followed. Assumes blip and treatment-free models correctly specified.

This function may be used in a similar fashion to more traditional modeling commands (such as lm). Users are referred to the primary 'DTRreg()' and 'DTRSurv()' help command (and associated literature) for information concerning model specification. In particular, we note that the predict function assumes that the treatment-free model has been correctly specified, as the treatment-free parameters are used in the prediction process.

Usage

## S3 method for class 'DTRreg'
predict(object, newdata, treat.range = NULL, ...)

Arguments

object

A model object generated by the function 'DTRreg()' or 'DWSurv()'.

newdata

A dataset (usually the data analyzed by DTRreg for which predicted outcomes are desired. If a new dataset is provided, variable names should correspond to those presented to 'DTRreg()' or 'DWSurv()'.

treat.range

If treatment is continuous (rather than binary), a vectors of the form c(min,max) which specify the minimum and maximum value the treatment may take at stage 1. If unspecified, this will be inferred from the treat.range provided with use of the original DTRreg command. As such, if no treatment range was specified there either, treat.range will be the minimum and maximum observed first stage treatment.

...

Space for additional arguments (not currently used)

Value

An n x 1 matrix of predicted outcome values.

Author(s)

Michael Wallace

References

Chakraborty, B., Moodie, E. E. M. (2013) Statistical Methods for Dynamic Treatment Regimes. New York: Springer.

Robins, J. M. (2004) Optimal structural nested models for optimal sequential decisions. In Proceedings of the Second Seattle Symposium on Biostatistics, D. Y. Lin and P. J. Heagerty (eds), 189-326. New York: Springer.

Wallace, M. P., Moodie, E. M. (2015) Doubly-Robust Dynamic Treatment Regimen Estimation Via Weighted Least Squares. Biometrics 71(3), 636-644 (doi:10.1111/biom.12306.)

Examples

# example single run of a 2-stage g-estimation analysis

set.seed(1)

# expit function
expit <- function(x) { 1.0 / (1.0 + exp(-x)) }

# sample size
n <- 10000

# variables (X = patient information, A = treatment)
X1 <- rnorm(n)
A1 <- rbinom(n, 1, expit(X1))
X2 <- rnorm(n)
A2 <- rbinom(n, 1, expit(X2))

# blip functions
gamma1 <- A1 * (1 + X1)
gamma2 <- A2 * (1 + X2)

# observed outcome: treatment-free outcome plus blip functions
Y <- exp(X1) + exp(X2) + gamma1 + gamma2 + rnorm(n)

# models to be passed to DTRreg
# blip model
blip.mod <- list(~ X1, ~ X2)
# treatment model (correctly specified)
treat.mod <- list(A1 ~ X1, A2 ~ 1)
# treatment-free model (incorrectly specified)
tf.mod <- list(~ X1, ~ X2)

# perform G-estimation
mod1 <- DTRreg(twoStageCont$Y, blip.mod, treat.mod, tf.mod, 
               data = twoStageCont, method = "gest")
  
# predicted Y for optimal treatment
dat <- data.frame(X1, X2, A1, A2)
predict(mod1, newdata = dat)