predict.bayesCureModel: Predict method.

In bayesCureRateModel: Bayesian Cure Rate Modeling for Time-to-Event Data

Predict method.

Description

Returns MAP estimates of the survival function and the conditional cured probability for a given set of covariates.

Usage

## S3 method for class 'bayesCureModel'
predict(object, newdata = NULL, tau_values = NULL, 
	burn = NULL, K_max = 1, alpha = 0.1, nDigits = 3, verbose = FALSE, ...)

Arguments

object	An object of class bayesCureModel
newdata	A data.frame with new data for the covariates. The column names as well as the class of each column (variable) should match with the input data.
tau_values	A vector of values for the response variable (time) for returning predictions for each row in the newdata.
burn	Positive integer corresponding to the number of mcmc iterations to discard as burn-in period
K_max	Maximum number of components in order to cluster the (univariate) values of the joint posterior distribution across the MCMC run. Used to identify the main mode of the posterior distribution.
alpha	Scalar between 0 and 1 corresponding to 1 - confidencel level for computing Highest Density Intervals. If set to NULL, the confidence intervals are not computed.
nDigits	A positive integer for printing the output, after rounding to the corresponding number of digits. Default: nDigits = 3.
verbose	Boolean. If set to TRUE (default) the function prints a summary of the predictions.
...	ignored.

Details

The values of the posterior draws are clustered according to a (univariate) normal mixture model, and the main mode corresponds to the cluster with the largest mean. The maximum number of mixture components corresponds to the K_max argument. The mclust library is used for this purpose. The inference for the latent cure status of each (censored) observation is based on the MCMC draws corresponding to the main mode of the posterior distribution. The FDR is controlled according to the technique proposed in Papastamoulis and Rattray (2018).

In case where covariate_levels is set to TRUE, the summary function also returns a list named p_cured_output with the following entries

mcmc

It is returned only in the case where the argument covariate_values is not NULL. A vector of posterior cured probabilities for the given values in covariate_values, per retained MCMC draw.

map

It is returned only in the case where the argument covariate_values is not NULL. The cured probabilities computed at the MAP estimate of the parameters, for the given values covariate_values.

tau_values

tau values

covariate_levels

covariate levels

index_of_main_mode

the subset of MCMC draws allocated to the main mode of the posterior distribution.

Value

A list with the following entries

map_estimate	Maximum A Posteriori (MAP) estimate of the parameters of the model.
highest_density_intervals	Highest Density Interval per parameter
latent_cured_status	Estimated posterior probabilities of the latent cure status per censored subject.
cured_at_given_FDR	Classification as cured or not, at given FDR level.
p_cured_output	It is returned only in the case where the argument covariate_values is not NULL. See details.
main_mode_index	The retained MCMC iterations which correspond to the main mode of the posterior distribution.

Author(s)

Panagiotis Papastamoulis

References

Papastamoulis and Milienos (2024). Bayesian inference and cure rate modeling for event history data. TEST doi: 10.1007/s11749-024-00942-w.

Papastamoulis and Rattray (2018). A Bayesian Model Selection Approach for Identifying Differentially Expressed Transcripts from RNA Sequencing Data, Journal of the Royal Statistical Society Series C: Applied Statistics, Volume 67, Issue 1.

Scrucca L, Fraley C, Murphy TB, Raftery AE (2023). Model-Based Clustering, Classification, and Density Estimation Using mclust in R. Chapman and Hall/CRC. ISBN 978-1032234953

See Also

cure_rate_MC3

Examples

# simulate toy data just for cran-check purposes        
	set.seed(10)
        n = 4
        # censoring indicators
        stat = rbinom(n, size = 1, prob = 0.5)
        # covariates
        x <- matrix(rnorm(2*n), n, 2)
        # observed response variable 
        y <- rexp(n)
#	define a data frame with the response and the covariates        
        my_data_frame <- data.frame(y, stat, x1 = x[,1], x2 = x[,2])
# run a weibull model with default prior setup
# considering 2 heated chains 
	fit1 <- cure_rate_MC3(survival::Surv(y, stat) ~ x1 + x2, data = my_data_frame, 
		promotion_time = list(distribution = 'exponential'),
		nChains = 2, 
		nCores = 1, 
		mcmc_cycles = 3, sweep=2)
	newdata <- data.frame(x1 = c(0.2,-1), x2 = c(-1,0))
	# return predicted values at tau = c(0.5, 1)
	my_prediction <- predict(fit1, newdata = newdata, 
		burn = 0, tau_values = c(0.5, 1))

bayesCureRateModel documentation built on Oct. 4, 2024, 1:07 a.m.