CERFIT (version 0.1.1)

predict.CERFIT: Get predictions from a CERFIT object

Description

Get predictions from a CERFIT object

Usage

# S3 method for CERFIT
predict(
  object,
  newdata = NULL,
  gridval = NULL,
  prediction = c("overall", "by iter"),
  type = c("response", "ITE", "node", "opT"),
  alpha = 0.5,
  ...
)

Value

The return value depends of the type argument. If type is response the function will return a matrix with n rows and the number of columns equal to the level of treatment. If type is ITE then it returns a matrix with n rows and a number of columns equal to one minus the levels of treatment. And if type is opT then it returns a matrix with n rows and two columns. With the first column denoting the optimal treatment and the second column denoting the optimal response.

Arguments

object

A fitted CERFIT object

newdata

New data to make predictions from. IF not provided will make predictions on training data

gridval

For continuous treatment. Controls for what values of treatment to predict

prediction

Return prediction using all trees ("overall") or using first i trees ("by iter")

type

Choose what value you wish to predict. Response will predict the response. ITE will predict the Individualized treatment effect. Node will predict the node. And opT will predict the optimal treatment for each observation.

alpha

For continuous treatment it is the mixing parameter for the elastic net regularization in each node. When equal to 0 it is ridge regression and when equal to 1 it is lasso regression.

...

Additional Arguments

Examples

fit <- CERFIT(Result_of_Treatment ~ sex + age + Number_of_Warts + Area + Time + Type | treatment,
data = warts,
ntrees = 30,
method = "RCT",
mtry = 2)
ite <- predict(fit,type = "ITE")