pirate (version 1.0.0)

gem_test: Implement Fitted GEM criterior on a Data Set

Description

Calculates the value of the treatment decision based on the information from a fitted GEM model.The information is contained in the gemObject, which is obtained from the gem_fit function. With gem_test_sample the value of decision based on a GEM model is obtained for a test sample that for each subject has observed outcome under only one treatment condition (this would be the situation when the test sample is a "real" data set). With gem_test_simsample the value of decision is calculated when the test sample has the outcome under all treatment conditions for all subjects (this would be the situation when simulated data is used).

Usage

gem_test_sample(dat, gemObject)
gem_test_simsample(y0, y1, XFrame, gemObject)

Arguments

dat

Data frame with first column as the treatment index, second column as outcome, and the remaining columns as covariates design matrix. The treatment index could only have two values ana the outcome should be of continuous type.

gemObject

A list containing the fitted GEM information, which could be the second element gemObject of the ouput from the gem_fit function or a list with the same structure.

y0

Outcome vector for all subjects under the first treatment assignment

y1

Outcome vector for all subjects under the second treatment assignment

XFrame

Design matrix of the predictors in the simulated sample with known outcomes under both treatment conditions

Value

PAB_gem Population average benefit of a the treatment regime based on a GEM modelPAB_unres Population average benefit of a treatment regime based on an unrestricted linear modelopt_gem The optimal treatment assignment for each subject

Details

The treatment decision rule estimated by the gem_fit function can be applied to a new (real) data set to estimate its value. It can also be applied to a simulated data set, where the outcome is known under both conditions, to study its performance of such treatment decision rule. These two functions correspond to those two situations and compute the population average benefit (or the value of the decision rule based on a GEM model).

Examples

#constructing the covariance matrix
co <- matrix(0.2, 10, 10)
diag(co) <- 1
dataEx <- data_generator1(d = 0.3, R2 = 0.5, v2 = 1, n = 300, 
                         co = co, beta1 = rep(1,10),inter = c(0,0))
#fit the GEM
dat <- dataEx[[1]]
model_nu <- gem_fit(dat = dat, method = "nu")

#calculate the population average benefit in the data sample
gem_test_sample(dat,model_nu[[2]])

#calculate the population average benefit when outcome under both treatment conditions 
#is known, usually in a simulated sample
bigData <- data_generator3(n = 1000,co = co,bet =dataEx[[2]], inter = c(0,0)) 
gem_test_simsample(bigData[[1]],bigData[[2]],bigData[[3]],model_nu[[2]])