predict.mp_cv: Prediction for a Fitted "mp_cv" Object

In chenshengkuang/mpersonalized: An R Package for Personalized Medicine in Meta-analysis and Multiple Outcomes

Description

This function predicts optimal treatment of new subjects for a cross-validated mpersonalized model.

Usage

## S3 method for class 'mp_cv'
predict(object, newx = NULL, weight = NULL,
  overall_rec = TRUE, ...)

Arguments

object	A fitted "mp_cv" object returned by "mpersonalized_cv" function
newx	Covariate matrix of new patients. If not supplied, by default the prediction is for the original dataset in the "mp_cv" object. Prediction results will differ based on whether newx is provided or not. Similar to predict.mp.
weight	A weight vector for the overall recommendation, only needed when overall_rec = TRUE. By default, equal weights are assigned to each study/outcome.
overall_rec	A logical value. If overall_rec = TRUE, an overall recommendation will be provided as an weighted average of the optimal treatment from each individual study/outcome. Only useful when newx is provided.
...	not used

Value

A list object with two elements. Similar to the returned value of predict.mp, but now it only predicts for the optimal parameter penalty.

opt_treatment

If newx is provided, a recommendation matrix with each row denoting a subject and each column denoting a study/outcome; otherwise, each element is a list of vectors with each vector representing the optimal treatment for each study/outcome. If overall_rec = TRUE, the weighted overall recommended treatment will be further provided as well. If the overall recommened treatment is equal to 0.5, it means the weighted sum is equal for 0 and 1.

benefit_score

Benefit scores computed from g_1, …, g_K. Similar to structure of opt_treatment.

Examples

set.seed(123)
sim_dat = simulated_dataset(n = 200, problem = "meta-analysis")
Xlist = sim_dat$Xlist; Ylist = sim_dat$Ylist; Trtlist = sim_dat$Trtlist

# fit different rules with group lasso penalty
mp_cvmod_diff = mpersonalized_cv(problem = "meta-analysis",
                                 Xlist = Xlist, Ylist = Ylist, Trtlist = Trtlist,
                                 penalty = "GL", single_rule = FALSE)

newx = matrix(rnorm(100 * mp_cvmod_diff$number_covariates), nrow = 100)

# predict on newx
pred_new = predict(object = mp_cvmod_diff, newx = newx, overall_rec = TRUE)

# predict on old dataset
pred_old = predict(object = mp_cvmod_diff)
set.seed(NULL)

chenshengkuang/mpersonalized documentation built on May 28, 2019, 7:16 p.m.