Package {spsurv}

Type: Package
Title: Bernstein Polynomial Based Semiparametric Survival Analysis
Version: 1.1.0
Description: Semiparametric survival analysis based on Bernstein polynomials. 'spsurv' includes proportional hazards, proportional odds and accelerated failure time frameworks for right-censored data. RV Panaro (2020) <doi:10.48550/arXiv.2003.10548>.
License: GPL-3
Biarch: true
Depends: R (≥ 4.1.0), survival (≥ 2.44-1.1), coda (≥ 0.19-3)
Imports: methods, Rcpp (≥ 0.12.0), RcppParallel (≥ 5.0.1), rstan (≥ 2.18.1), generics, viridis
LinkingTo: BH (≥ 1.66.0), Rcpp (≥ 0.12.0), RcppEigen (≥ 0.3.3.3.0), RcppParallel (≥ 5.0.1), rstan (≥ 2.18.1), StanHeaders (≥ 2.18.0)
Suggests: testthat (≥ 2.1.0), knitr, rmarkdown (≥ 2.31), ggplot2 (≥ 3.3.0), KMsurv, parsnip (≥ 1.0.0), censored (≥ 0.3.0), workflows (≥ 1.1.0), tidybayes (≥ 3.0.0), posterior (≥ 1.4.0), covr, rsample (≥ 1.1.0), rsurv, rstantools (≥ 1.5.1)
VignetteBuilder: knitr
Encoding: UTF-8
SystemRequirements: GNU make
BugReports: https://github.com/rvpanaro/spsurv/issues
URL: https://github.com/rvpanaro/spsurv, https://rvpanaro.github.io/spsurv/
Config/roxygen2/version: 8.0.0
Config/rstantools/auto_config: FALSE
NeedsCompilation: yes
Packaged: 2026-09-28 07:43:22 UTC; rvpanaro
Author: Renato Panaro ORCID iD [aut, cre, cph] (URL: https://rvpanaro.github.io/), Fábio Demarqui ORCID iD [ths, rev] (URL: http://www.est.ufmg.br/~fndemarqui/), Vinicius Mayrink ORCID iD [ths] (URL: http://www.est.ufmg.br/~vdinizm/)
Maintainer: Renato Panaro <rvpanaro@gmail.com>
Repository: CRAN
Date/Publication: 2026-09-28 09:00:02 UTC

The 'spsurv' package.

Description

A set of flexible routines to allow semiparametric survival regression modeling based on Bernstein polynomial, including Bernstein PH model (BPPH), Bernstein PO model (BPPO), and Bernstein AFT model (BPAFT) for right-censored data.

Details

spbp fits semi-parametric models for time-to-event survival data. Non-informative right-censoring assumption is available. Any user-defined Bernstein polynomial can be user-defined using an arbitrary degree, i.e. highest basis polynomials order.

The framework takes advantage of fully likelihood methods since the polynomial parameters are used to estimate the baseline functions. Even so, this is said to be semi-parametric since this approach does not rely on any distribution. Unlike the Cox model, the BP based models provide smooth hazard and survival curve estimates. For Bayesian fits, users should routinely inspect divergences, split-\hat R, and effective sample sizes, and tighten NUTS controls (e.g., increasing adapt_delta, iter, and warmup) when warnings are present.

Value

none

Author(s)

rvpanaro@gmail.com

References

Panaro R.V. (2020). spsurv: An R package for semi-parametric survival analysis. arXiv preprint arXiv:2003.10548.

Demarqui, F. N., & Mayrink, V. D. (2019). A fully likelihood-based approach to model survival data with crossing survival curves. arXiv preprint arXiv:1910.02406.

Demarqui, F. N., Mayrink, V. D., & Ghosh, S. K. (2019). An Unified Semiparametric Approach to Model Lifetime Data with Crossing Survival Curves. arXiv preprint arXiv:1910.04475.

Osman, M., & Ghosh, S. K. (2012). Nonparametric regression models for right-censored data using Bernstein polynomials. Computational Statistics & Data Analysis, 56(3), 559-573.

Lorentz, G. G. (1953). Bernstein polynomials. American Mathematical Society.

Internal: Calculate the posterior mode

Description

Internal: Calculate the posterior mode

Usage

.mode(ext)

Internal: AFT basis evaluator aligned with Stan transformed parameters

Description

Internal: AFT basis evaluator aligned with Stan transformed parameters

Usage

.spbp_aft_basis(time_aft, tau_a, tau_b, P)

Arguments

time_aft

Numeric vector of AFT residual-scale times y = log(t) - eta.

tau_a

Lower bound used by the fitted model.

tau_b

Upper bound used by the fitted model.

P

Power-basis transformation matrix from pw.basis().

Value

list(g = ..., G = ...) where g is hazard basis and G is cumulative-hazard basis on the AFT residual scale.

Bayesian fit for spbp (internal)

Description

Bayesian fit for spbp (internal)

Usage

.spbp_bayes(standata, hessian = FALSE, verbose = FALSE, chains = 1, ...)

Internal: compute survival CI bands (survfit-style)

Description

Internal: compute survival CI bands (survfit-style)

Usage

.survfit_confint(
  p,
  se,
  logse = TRUE,
  conf.type,
  conf.int,
  selow,
  ulimit = TRUE
)

Akaike information criterion for fitted spbp models

Description

Akaike information criterion for fitted spbp models

Usage

## S3 method for class 'spbp'
AIC(object, ..., k = 2)

Arguments

object

A fitted "spbp" object.

...

Additional fitted models for comparison.

k

Penalty per parameter (default 2).

Analysis of deviance table for nested spbp models

Description

Compare nested MLE fits. With one model, builds a sequential table by refitting intercept-only and nested submodels along the formula term order (same idea as anova(fit) for survreg). With multiple models, compares each consecutive pair (same convention as anova(fit0, fit1) for survreg).

Usage

## S3 method for class 'spbp'
anova(object, ..., test = "Chisq")

Arguments

object

A fitted "spbp" object.

...

Additional nested fitted models (MLE, same family recommended).

test

Which test to report (only "Chisq" is supported).

Value

An "anova" object (also a data.frame). Pairwise comparisons use Terms, Resid. Df, -2*LL, Test, Df, Deviance, and Pr(>Chi) columns (as in survival::survreg). Sequential single-model tables use Df, Deviance, Resid. Df, -2*LL, and Pr(>Chi) with row names NULL and term labels.

Convert spbp Bayesian fit to posterior draws

Description

Convert spbp Bayesian fit to posterior draws

Usage

as_draws_df.spbp(x, variables = c("beta", "gamma"), ...)

Arguments

x

A fitted "spbp" object from approach = "bayes".

variables

Character vector of posterior components to include ("beta", "gamma", "log_lik").

...

Not used.

Value

A posterior::draws_df object on the back-transformed scale.

Augment data with spbp model information

Description

Add fitted values, residuals, and optional survival predictions to the training (or supplied) data.

Usage

## S3 method for class 'spbp'
augment(
  x,
  data = NULL,
  eval_time = NULL,
  type = c("martingale", "deviance", "cox-snell", "coxsnell"),
  ...
)

Arguments

x

A fitted "spbp" object.

data

Optional data frame; defaults to the training data when available.

eval_time

Optional numeric vector of times for nested survival predictions (same structure as predict(x, type = "survival")).

type

Residual type passed to residuals.spbp.

...

Not used.

Value

A data.frame with original rows plus .residual and, when eval_time is set, a list-column .pred.

Bernstein polynomial baseline specification

Description

Helper for specifying a Bernstein polynomial baseline via the dist / baseline arguments accepted by spbp.

Usage

bernstein(m = NULL)

Arguments

m

Bernstein polynomial degree (number of basis coefficients).

Value

A list with components baseline and m.

Examples

bernstein(5)

Bernstein basis polynomials calculations

Description

Bernstein basis polynomials calculations

Usage

bp.basis(time, degree, tau = max(time))

Arguments

time

a vector of times.

degree

Bernstein polynomial degree

tau

must be greater than times maximum value observed.

Value

A list containing matrices g and G corresponding BP basis and corresponding tau value used to compute them.

Bernstein survival regression specification

Description

Convenience constructor mapping family to the appropriate parsnip model (proportional_hazards, proportional_odds, or survival_reg).

Usage

bp_survival_reg(
  family = c("ph", "po", "aft"),
  mode = "censored regression",
  engine = "spsurv"
)

Arguments

family

Model family: "ph", "po", or "aft".

mode

Model mode (default "censored regression").

engine

parsnip engine (default "spsurv" for MLE).

Value

A parsnip model specification.

Bernstein AFT Model

Description

Fits the BPAFT model to time-to-event data.

Usage

bpaft(
  formula,
  degree = NULL,
  data,
  approach = c("mle", "bayes"),
  dist = NULL,
  baseline = NULL,
  ...
)

Arguments

formula

a Surv object with time to event observations, right censoring status and explanatory terms.

degree

Bernstein polynomial degree. If omitted, defaults to ceiling(sqrt(n)) for n = nrow(data) (see spbp).

data

a data.frame object.

approach

Bayesian or maximum likelihood estimation methods, default is approach = "mle".

dist

optional baseline specification; use bernstein(m) for the Bernstein polynomial degree.

baseline

optional alias for dist.

...

further arguments passed to or from other methods

Value

An object of class spbp, including component degree.

See Also

spbp, bpph and bppo for other BP based models.

Examples


library("spsurv")
data("veteran", package = "survival")

fit <- bpaft(Surv(time, status) ~ karno + celltype,
  data = veteran
)

summary(fit)

Bernstein PH Model

Description

Fits the BPPH model to time-to-event data.

Usage

bpph(
  formula,
  degree = NULL,
  data,
  approach = c("mle", "bayes"),
  dist = NULL,
  baseline = NULL,
  ...
)

Arguments

formula

a Surv object with time to event observations, right censoring status and explanatory terms.

degree

Bernstein polynomial degree. If omitted, defaults to ceiling(sqrt(n)) for n = nrow(data) (see spbp).

data

a data.frame object.

approach

Bayesian or maximum likelihood estimation methods, default is approach = "mle".

dist

optional baseline specification; use bernstein(m) for the Bernstein polynomial degree.

baseline

optional alias for dist.

...

further arguments passed to or from other methods

Value

An object of class spbp, including component degree.

See Also

spbp, bppo and bpaft for other BP based models.

Examples


library("spsurv")
data("veteran", package = "survival")

fit <- bpph(Surv(time, status) ~ karno + factor(celltype),
  data = veteran
)

summary(fit)

Bernstein PO Model

Description

Fits the BPPO model to time-to-event data.

Usage

bppo(
  formula,
  degree = NULL,
  data,
  approach = c("mle", "bayes"),
  dist = NULL,
  baseline = NULL,
  ...
)

Arguments

formula

a Surv object with time-to-event observations, right censoring status and explanatory terms.

degree

Bernstein polynomial degree. If omitted, defaults to ceiling(sqrt(n)) for n = nrow(data) (see spbp).

data

a data.frame object.

approach

Bayesian or maximum likelihood estimation methods, default is approach = "mle".

dist

optional baseline specification; use bernstein(m) for the Bernstein polynomial degree.

baseline

optional alias for dist.

...

further arguments passed to or from other methods

Value

An object of class spbp, including component degree.

See Also

spbp, bpph and bpaft for other BP based models.

Examples


library("spsurv")
data("veteran", package = "survival")

fit <- bppo(Surv(time, status) ~ karno + celltype,
  data = veteran
)

summary(fit)

Estimated regression coefficients

Description

Estimated regression coefficients

Usage

## S3 method for class 'spbp'
coef(object, summary = c("mean", "median", "mode"), ...)

Arguments

object

an object of the class spbp

summary

posterior summary if method = "bayes" in x

...

further arguments passed to or from other methods

Value

the estimated regression coefficients

Confidence intervals for the regression coefficients

Description

Confidence intervals for the regression coefficients

Usage

## S3 method for class 'spbp'
confint(object, parm, level = 0.95, ...)

Arguments

object

a fitted model object.

parm

a specification of which parameters are to be given confidence intervals: regression coefficient names and/or Bernstein baseline names (e.g. "gamma[1]"). If missing, all regression coefficients are used.

level

the confidence level required.

...

further arguments passed to parent method

Value

100(1-alpha) confidence intervals for the requested parameters.

Generic S3 method credint

Description

Generic S3 method credint

Usage

credint(x, ...)

Arguments

x

a fitted model object

...

further arguments passed to parent method

Value

Generic function dispatching to credint.spbp.

Credible intervals for the regression coefficients

Description

Credible intervals for the regression coefficients

Usage

## S3 method for class 'spbp'
credint(x, prob = 0.95, type = c("HPD", "Equal-Tailed"), ...)

Arguments

x

an object of the class x.

prob

the probability level required.

type

interval type.

...

further arguments passed to or from other methods.

Value

100(1-alpha) credible intervals for the regression coefficients

Parameter estimates for fitted spbp models

Description

Parameter estimates for fitted spbp models

Usage

estimates(object, ...)

Arguments

object

A fitted "spbp" object.

...

Not used.

Value

Named numeric vector of regression and Bernstein-polynomial parameters.

Extract AIC from a fitted spbp model

Description

Extract AIC from a fitted spbp model

Usage

## S3 method for class 'spbp'
extractAIC(fit, scale = 0, k = 2, ...)

Arguments

fit

A fitted "spbp" object.

scale

Not used.

k

Penalty per parameter (default 2).

...

Not used.

ggplot2 residual diagnostic plots for spbp models

Description

Convenience wrapper around residuals.spbp that returns a ggplot2 object for martingale, deviance, or Cox-Snell residuals.

Usage

ggresiduals(
  object,
  type = c("martingale", "deviance", "cox-snell", "coxsnell"),
  against = c("fitted", "index"),
  ...
)

Arguments

object

A fitted "spbp" object.

type

Residual type: "martingale" (default), "deviance", "cox-snell", or "coxsnell".

against

What to plot on the x-axis: "fitted" (default, Cox-Snell cumulative hazard) or "index" (observation index).

...

Further arguments passed to residuals.spbp.

Value

A ggplot object (requires ggplot2).

Examples

## Not run: 
library(spsurv)
library(ggplot2)
data(veteran, package = "survival")
fit <- bpph(Surv(time, status) ~ karno, data = veteran, degree = 4)
ggresiduals(fit, type = "martingale")

## End(Not run)

Glance at a fitted spbp model

Description

Return a one-row broom/generics-style summary of model-level statistics.

Usage

## S3 method for class 'spbp'
glance(x, ...)

Arguments

x

A fitted "spbp" object.

...

Currently unused.

Value

A one-row data.frame with model-level statistics. Columns that are entirely NA for the fitted object are omitted.

Log-likelihood for fitted spbp models

Description

Log-likelihood for fitted spbp models

Usage

## S3 method for class 'spbp'
logLik(object, ...)

Arguments

object

A fitted "spbp" object.

...

Not used.

Model.matrix method for fitted spbp models

Description

Model.matrix of a fitted spbp model.

Usage

## S3 method for class 'spbp'
model.matrix(object, ...)

Arguments

object

an object of class 'spbp', see spbp.

...

arguments passed to parent method.

Value

The model matrix.

See Also

spbp, model.matrix

Examples


library("spsurv")
data("veteran", package = "survival")

fit <- bpph(Surv(time, status) ~ karno + factor(celltype),
  data = veteran
)

model.matrix(fit)

BP based models plot.

Description

Plot for a fitted spbp model.

Usage

## S3 method for class 'spbp'
plot(
  x,
  main,
  graph = c("baseline", "basis"),
  cumulative = FALSE,
  frame = FALSE,
  lwd = 3,
  ...
)

Arguments

x

an object of class 'spbp' result of a spbp fit.

main

graph title

graph

type of polynomial graph, default is "basis"

cumulative

TRUE for odds and cumulative hazard

frame

graphical parameter; default is FALSE

lwd

graphical parameter; default is 3

...

further arguments passed to or from other methods

See Also

spbp.

Examples


library("spsurv")
data("veteran", package = "survival")

fit <- bpph(Surv(time, status) ~ karno + factor(celltype),
  data = veteran
)
plot(fit)

Predicted survival as a tidy data frame

Description

Survival (and optional CI) on a time grid, as a data.frame for ggplot2::geom_line.

Usage

## S3 method for class 'spbp'
predict(
  object,
  newdata = NULL,
  times = NULL,
  eval_time = NULL,
  type = NULL,
  conf_type = NULL,
  interval = 0.95,
  interval.type = c("hpd", "quantile"),
  monotone = NULL,
  ...
)

Arguments

object

Fitted "spbp" from spbp / bpph / etc.

newdata

Optional data.frame of covariate profiles (same convention as survfit.spbp).

times

Time grid. Default: a dense sequence from 0 to the maximum observed time (suitable for smooth ggplot2::geom_line / geom_ribbon plots). Pass a long seq(...) to override resolution; use observed event times only if stepwise curves are intended.

eval_time

Evaluation times for type = "survival" (required).

type

Prediction type. For tidymodels/censored compatibility use "survival", "time", or "linear_pred". For survival curves (default), use NULL, "curve", or a confidence transformation ("log", "log-log", "plain").

conf_type

Confidence interval transformation for curve predictions ("log", "log-log", "plain"). Used when type is NULL, "curve", or a confidence transformation name.

interval

Confidence level for curve predictions; passed to survfit.spbp.

interval.type, monotone

Passed to survfit.spbp (Bayesian fits).

...

Passed to survfit.spbp for curve predictions.

Value

For curve predictions, same structure as as.data.frame.survfitbp. For type = "survival", a tibble with list-column .pred (elements contain .eval_time and .pred_survival). For type = "time", a tibble with .pred_time. For type = "linear_pred", a tibble with .pred_linear_pred.

See Also

survfit.spbp, augment.spbp

Examples

data(veteran, package = "survival")
fit <- bpph(Surv(time, status) ~ karno, data = veteran, approach = "mle", init = 0)
pr <- predict(fit, times = seq(0, 400, by = 2))
predict(fit, veteran[1:2, ], type = "survival", eval_time = c(100, 200))
predict(fit, veteran[1:2, ], type = "time")
## Not run: 
  ggplot2::ggplot(pr, ggplot2::aes(time, surv)) + ggplot2::geom_line()

## End(Not run)

Bernstein Polynomial Based Regression Object Print

Description

Bernstein Polynomial Based Regression Object Print

Usage

## S3 method for class 'spbp'
print(
  x,
  bp.param = FALSE,
  show_baseline = FALSE,
  digits = 2,
  signif.stars = getOption("show.signif.stars"),
  what = "summary",
  ...
)

Arguments

x

an object of class spbp.

bp.param

print BP parameters only (alias for baseline-only display).

show_baseline

logical; append Bernstein baseline Wald table after regression coefficients (MLE only). See also print(fit, bp.param = TRUE) for baseline-only output.

digits

number of digits to display; defaults to 2.

signif.stars

see getOption.

what

character vector; any of "summary", "tidy", and "glance" for broom-style coefficient and model-level tables.

...

further arguments passed to summary.spbp.

Value

none

Bernstein Polynomial Based Regression Object Summary BPAFT Bayes

Description

Bernstein Polynomial Based Regression Object Summary BPAFT Bayes

Usage

## S3 method for class 'summary.bpaft.bayes'
print(...)

Arguments

...

further arguments passed to or from other methods

Value

none

Bernstein Polynomial Based Regression Object Summary BPAFT MLE

Description

Bernstein Polynomial Based Regression Object Summary BPAFT MLE

Usage

## S3 method for class 'summary.bpaft.mle'
print(...)

Arguments

...

further arguments passed to or from other methods

Value

none

Bernstein Polynomial Based Regression Object Summary BPPH Bayes

Description

Bernstein Polynomial Based Regression Object Summary BPPH Bayes

Usage

## S3 method for class 'summary.bpph.bayes'
print(...)

Arguments

...

further arguments passed to or from other methods

Value

none

Bernstein Polynomial Based Regression Object Summary BPPH MLE

Description

Bernstein Polynomial Based Regression Object Summary BPPH MLE

Usage

## S3 method for class 'summary.bpph.mle'
print(...)

Arguments

...

further arguments passed to or from other methods

Value

none

Bernstein Polynomial Based Regression Object Summary BPPO Bayes

Description

Bernstein Polynomial Based Regression Object Summary BPPO Bayes

Usage

## S3 method for class 'summary.bppo.bayes'
print(...)

Arguments

...

further arguments passed to or from other methods

Value

none

Bernstein Polynomial Based Regression Object BPPO MLE

Description

Bernstein Polynomial Based Regression Object BPPO MLE

Usage

## S3 method for class 'summary.bppo.mle'
print(...)

Arguments

...

further arguments passed to or from other methods

Value

none

Bernstein Polynomial Based Regression Object Summary Bayes

Description

Bernstein Polynomial Based Regression Object Summary Bayes

Usage

## S3 method for class 'summary.spbp.bayes'
print(x, digits = 2, signif.stars = getOption("show.signif.stars"), ...)

Arguments

x

a summary.spbp.bayes object

digits

number of digits to display.

signif.stars

see getOption

...

further arguments passed to or from other methods

Value

none

Bernstein Polynomial Based Regression Object Summary MLE

Description

Bernstein Polynomial Based Regression Object Summary MLE

Usage

## S3 method for class 'summary.spbp.mle'
print(x, digits = 2, signif.stars = getOption("show.signif.stars"), ...)

Arguments

x

a summary.spbp.mle object

digits

number of digits to display.

signif.stars

see getOption

...

further arguments passed to or from other methods

Value

none

Proportional odds regression model specification

Description

Proportional odds survival regression with a Bernstein polynomial baseline (bppo). This model type is registered by spsurv for use with tidymodels; it is not part of the censored package.

Usage

proportional_odds(mode = "censored regression", engine = "spsurv")

Arguments

mode

A single character string for the prediction outcome mode. The only possible value for this model is "censored regression".

engine

A single character string specifying what computational engine to use for fitting.

Value

A model specification.

Power basis polynomials calculations

Description

Power basis polynomials calculations

Usage

pw.basis(degree)

Arguments

degree

Bernstein polynomial degree

Value

A list containing matrices g and G corresponding BP basis and corresponding tau value used to compute them.

Rank fitted spbp models by AIC

Description

Convenience wrapper around AIC that returns models sorted from lowest to highest AIC.

Usage

rank_models(...)

Arguments

...

One or more fitted "spbp" objects (MLE).

Value

A data.frame with columns fit, model, degree, aic, and npars.

Examples

## Not run: 
library(spsurv)
data(veteran, package = "survival")
f1 <- bpph(Surv(time, status) ~ 1, data = veteran, degree = 4)
f2 <- bpph(Surv(time, status) ~ karno, data = veteran, degree = 4)
rank_models(f1, f2)

## End(Not run)

BP based models residuals.

Description

Residuals for a fitted spbp model.

Usage

## S3 method for class 'spbp'
residuals(
  object,
  type = c("martingale", "deviance", "cox-snell", "coxsnell"),
  ...
)

Arguments

object

an object of class 'spbp' result of a spbp fit.

type

type of residuals; default is "martingale". Also "deviance", "cox-snell", and "coxsnell" (alias).

...

arguments passed to parent method.

See Also

spbp, spbp.

Examples


library("spsurv")
data("veteran", package = "survival")

fit <- bpph(Surv(time, status) ~ karno + factor(celltype),
  data = veteran
)

residuals(fit)

Standard errors for fitted spbp model parameters

Description

Standard errors for fitted spbp model parameters

Usage

se(object, ...)

Arguments

object

A fitted "spbp" object.

...

Not used.

Value

Named numeric vector of standard errors aligned with estimates.

spbp: The BP Based Survival Analysis Function

Description

Semiparametric Survival Analysis Using Bernstein Polynomial

Usage

spbp(formula, ...)

Arguments

formula

a Surv response with event time, censoring status, and optional covariates.

...

Arguments passed to spbp.default, including data, model, approach, and degree. Further arguments in ... are passed to rstan::optimizing (MLE) or rstan::sampling (Bayes), e.g. iter, chains, init.

Details

Fits Bernstein PH, PO, or AFT models to survival data via Stan (MLE or Bayesian).

The generic dispatches to spbp.default for formula objects. Convenience wrappers bpph, bppo, and bpaft fix the model family. See vignette("getting-started", package = "spsurv") for a tutorial, vignette("model-families", package = "spsurv") for PH / PO / AFT comparison, and vignette("bp-degree", package = "spsurv") for choosing the Bernstein polynomial degree.

Value

An object of class "spbp". See spbp.default for the list of components (coefficients, bp.param, degree, etc.).

See Also

spbp.default, bpph, bppo, bpaft, bernstein

Examples


library("spsurv")
data("veteran", package = "survival")

fit_mle <- spbp(Surv(time, status) ~ karno + factor(celltype),
  data = veteran, model = "po"
)
summary(fit_mle)

fit_bayes <- spbp(Surv(time, status) ~ karno + factor(celltype),
  data = veteran, model = "po", approach = "bayes",
  cores = 1, iter = 300, chains = 1,
  priors = list(
    beta = c("normal(0,5)"),
    gamma = "halfnormal(0,5)"
  )
)

summary(fit_bayes)

spbp: The BP Based Semiparametric Survival Analysis Function

Description

spbp: The BP Based Semiparametric Survival Analysis Function

Usage

## Default S3 method:
spbp(
  formula,
  degree,
  data,
  approach = c("mle", "bayes"),
  model = c("ph", "po", "aft"),
  priors = list(beta = c("normal(0,2)"), gamma = c("lognormal(0,4)"), frailty =
    c("gamma(0.01,0.01)")),
  cores = .spbp_default_cores(),
  scale = TRUE,
  dist = NULL,
  baseline = NULL,
  verbose = FALSE,
  chains = 4,
  ...
)

Arguments

formula

a Surv object with time to event, status and explanatory terms

degree

Bernstein polynomial degree (integer). If omitted and neither dist nor baseline supplies bernstein(m), the default is ceiling(sqrt(n)) where n is the number of rows in data.

data

a data.frame object

approach

Bayesian or Maximum Likelihood estimation methods; default is "mle"

model

Bernstein PH ("ph"), PO ("po"), or AFT ("aft") model; default is "ph"

priors

prior settings for the Bayesian approach; 'normal' or 'cauchy' for beta; 'lognormal' or 'loglogistic' for gamma (BP coefficients). Defaults are normal(0,2) for standardised regression coefficients and lognormal(0,4) for Bernstein coefficients.

cores

number of core threads to use (Bayes sampling)

scale

logical; indicates whether to center and scale the data

dist

optional baseline specification; use bernstein(m) for the Bernstein polynomial degree

baseline

optional alias for dist

verbose

passed to Stan

chains

number of MCMC chains (Bayes)

...

further arguments passed to rstan::optimizing (MLE) or rstan::sampling (Bayes), e.g. iter, warmup, init.

Details

Right-censored survival data are modeled with a Bernstein-polynomial baseline and regression on covariates. With approach = "mle", parameters are estimated by Stan's optimizer and approximate inference uses the Hessian when available. With approach = "bayes", posterior samples are drawn with NUTS; use summary, tidy.spbp, and glance.spbp for output. Covariates are centered and scaled when scale = TRUE (default).

The returned object includes:

coefficients

Regression estimates on the original covariate scale.

bp.param

Bernstein baseline coefficients (gamma).

degree

Polynomial degree used (also in call$degree).

loglik

MLE: intercept-only and full-model log-likelihoods; Bayes: posterior mean pointwise log-likelihoods.

call

Matched call with approach, model, and degree.

Value

An object of class spbp. Component degree records the Bernstein polynomial degree used in the fit (also stored in call$degree).

See Also

bpph, bppo, bpaft, bernstein, summary.spbp

Posterior survival curves in long tidy format

Description

Posterior survival curves in long tidy format

Usage

spread_surv_draws.spbp(object, times, newdata = NULL, ...)

Arguments

object

A Bayesian "spbp" fit.

times

Numeric vector of evaluation times.

newdata

Optional covariate profiles.

...

Not used.

Value

A data.frame with .chain, .iteration, .draw, time, surv, and optional covariate columns.

Fit proportional hazards model (spsurv parsnip engine)

Description

Fit proportional hazards model (spsurv parsnip engine)

Usage

spsurv_fit_proportional_hazards(formula, data, case_weights = NULL, ...)

Fit proportional odds model (spsurv parsnip engine)

Description

Fit proportional odds model (spsurv parsnip engine)

Usage

spsurv_fit_proportional_odds(formula, data, case_weights = NULL, ...)

Fit AFT model (spsurv parsnip engine)

Description

Fit AFT model (spsurv parsnip engine)

Usage

spsurv_fit_survival_reg(formula, data, case_weights = NULL, ...)

Predict linear predictor (spsurv parsnip engine)

Description

Predict linear predictor (spsurv parsnip engine)

Usage

spsurv_pred_linear_pred(object, new_data, ...)

Predict survival probabilities (spsurv parsnip engine)

Description

Predict survival probabilities (spsurv parsnip engine)

Usage

spsurv_pred_survival(object, new_data, eval_time, ...)

Predict median event time (spsurv parsnip engine)

Description

Predict median event time (spsurv parsnip engine)

Usage

spsurv_pred_time(object, new_data, ...)

Bernstein Polynomial Based Regression Object Summary

Description

Bernstein Polynomial Based Regression Object Summary

Usage

## S3 method for class 'spbp'
summary(
  object,
  interval = 0.95,
  compact = TRUE,
  show_call = TRUE,
  show_intervals = TRUE,
  show_baseline = FALSE,
  mle_test = c("lr", "wald"),
  bayes_criterion = c("waic", "dic", "lpml"),
  ...
)

Arguments

object

an object of class spbp

interval

interval coverage (confidence or credibility)

compact

logical; if TRUE, print.summary methods show essential output only.

show_call

logical; include the model call in printed summary.

show_intervals

logical; include interval table in printed summary.

show_baseline

logical; include Bernstein baseline (gamma) Wald table in printed summary (MLE only).

mle_test

global test to print for MLE summaries: "lr" or "wald".

bayes_criterion

global criterion to print for Bayesian summaries: "waic", "dic", or "lpml".

...

further arguments passed to or from other methods

Value

An object of class analogous to for e.g. 'summary.bppo.bayes'.

BP-based model survival curves

Description

Compute survival curves for a fitted spbp model.

Usage

## S3 method for class 'spbp'
survfit(
  formula,
  newdata = NULL,
  times = NULL,
  se.fit = TRUE,
  interval = 0.95,
  type = c("log", "log-log", "plain"),
  interval.type = c("hpd", "quantile"),
  monotone = NULL,
  tidy = FALSE,
  baseline = FALSE,
  ...
)

## S3 method for class 'survfitbp'
as.data.frame(x, row.names = NULL, optional = FALSE, ...)

## S3 method for class 'survfitbp'
plot(
  x,
  conf.int,
  col = 1,
  lty = 1,
  lwd = 1.5,
  xlab = "Time",
  ylab = "Survival probability",
  ylim = c(0, 1),
  mark.time = FALSE,
  ...
)

Arguments

formula

An object of class "spbp" returned by spbp.

newdata

Optional data frame used to obtain survival curves for specific covariate values.

times

Optional evaluation times: a Surv object (legacy), or a non-negative numeric vector (unique values are used; suitable for smooth ggplot2::geom_line plots via predict.spbp or as.data.frame.survfitbp).

se.fit

Logical; if TRUE, compute standard errors.

interval

Confidence level for intervals (e.g. 0.95).

type

Character; confidence interval transformation. One of "log", "log-log", or "plain".

interval.type

For Bayesian fits only: "hpd" (default) or "quantile" (equal-tailed).

monotone

Logical; for Bayesian fits, enforce non-increasing credible-band limits over time so ribbons plot smoothly with ggplot2::geom_line. Defaults to TRUE when approach = "bayes".

tidy

Logical; if TRUE, return a data.frame for ggplot2. instead of a "survfit" object.

baseline

Logical; if TRUE, return the baseline survival curve S_0(t) at observed event times (or times) with no covariate effect. When tidy = TRUE, returns a simple data.frame with columns time and surv.

...

Further arguments passed to survfit.coxph for the reference Cox object (e.g. conf.type is ignored; use type instead).

x

Object from survfit.spbp.

row.names, optional

Unused; included for S3 consistency.

conf.int

Logical; draw pointwise interval limits. Defaults to TRUE for a single curve when limits are present.

col, lty, lwd, xlab, ylab, ylim

Graphical parameters.

mark.time

Ignored (Bernstein curves are continuous).

Value

An object of class "survfit" (with classes survfitbp, survfitcox, survfit).

A data.frame with columns id (curve index), time, surv, lower, upper, cumhaz, std.err.

Functions

See Also

spbp, survfit, predict.spbp, survfit.spbp (as.data.frame method).

Examples

library(spsurv)
data(veteran, package = "survival")
fit <- bpph(Surv(time, status) ~ karno + factor(celltype), data = veteran)
survfit(fit)
plot(survfit(fit, times = seq(0, max(veteran$time), length.out = 80)))

data(veteran, package = "survival")
fit <- bpph(Surv(time, status) ~ karno, data = veteran, approach = "mle", init = 0)
sf <- survfit(fit, times = seq(0, max(veteran$time), length.out = 80))
head(as.data.frame(sf))

Tidy summary for fitted spbp models

Description

Create a broom/generics-style tidy data frame with one row per model term.

Usage

## S3 method for class 'spbp'
tidy(
  x,
  conf.int = FALSE,
  conf.level = 0.95,
  exponentiate = FALSE,
  component = c("coef", "baseline", "all"),
  ...
)

Arguments

x

A fitted "spbp" object.

conf.int

Logical; include confidence/credible interval columns.

conf.level

Interval coverage level.

exponentiate

Logical; if TRUE, exponentiate the estimate and interval columns (regression coefficients only).

component

Which parameters to return: "coef" (default), "baseline" (Bernstein gamma weights), or "all".

...

Currently unused.

Value

A data.frame with columns term, estimate, std.error, and when available statistic, p.value, conf.low, conf.high. Columns that are entirely NA are omitted.

Covariance of the regression coefficients

Description

Uses block-wise inversion of the negative Hessian, with a clear split between the regression coefficients (beta) and the Bernstein polynomial coefficients (gamma).

Usage

## S3 method for class 'spbp'
vcov(object, bp.param = FALSE, mask_unstable_gamma = TRUE, ...)

Arguments

object

an object of the class spbp

bp.param

return Bernstein Polynomial variance.

mask_unstable_gamma

when TRUE (default), set \gamma-related entries to NA if the Bernstein information block is ill-conditioned. Set FALSE internally when propagating survival uncertainty despite instability.

...

arguments passed to parent method.

Value

the variance-covariance matrix associated with the regression coefficients.