Bayesian Decision Curve Analysis for Survival outcomes

Usage

dca_surv(
  .data,
  prediction_time,
  thresholds = seq(0, 0.5, length = 51),
  keep_draws = TRUE,
  keep_fit = FALSE,
  summary_probs = c(0.025, 0.975),
  positivity_prior = c(1, 1),
  shape_prior = c("student", "gamma"),
  scale_prior = c("student", "gamma"),
  shape_prior_pars = c(10, 0, 1.5),
  scale_prior_pars = c(30, 0, 100),
  prior_only = FALSE,
  iter = 4000,
  refresh = 0,
  ...
)

Arguments

.data: dataframe whose first column named "outcomes" is a survival::Surv object and remaining columns are the decision strategies to assess.
prediction_time: Prediction time horizon (e.g., if models predict risk of death at one year and data is in year, prediction_time should be 1.)
thresholds: Decision thresholds -- within interval (0, 1).
keep_draws: If true, posterior draws are kept in the output object.
keep_fit: If true, stanfit object is kept in the output object.
summary_probs: Probabilities for posterior credible intervals (defaults to a 95% Cr.I.).
positivity_prior: Shape parameters for prior on positivity probability.
shape_prior: type of prior distribution for shape parameter of the Weibull distribution. Either "student" or "gamma".
scale_prior: type of prior distribution for scale parameter of the Weibull distribution. Either "student" or "gamma".
shape_prior_pars: vector with prior parameters for the prior shape of the Weibull distribution. If shape_prior="student", it should be a vector of length 3 with degrees of freedom, mean, and scale, respectively; if shape_prior="gamma", it should be a vector of length 2 with shape and rate, respectively.
prior_only: If TRUE, samples from the prior only.
iter: Passed to rstan::sampling. Number of iterations/draws for Stan.
refresh: Controls verbosity of rstan::sampling
...: Arguments passed to rstan::sampling (e.g. iter, chains). # nolint

Value

An object of class BayesDCASurv

Examples

if (FALSE) {
data(dca_survival_data)
fit <- dca_surv(dca_survival_data, prediction_time = 1, iter = 1000, chains = 1)
plot(fit)
}