Package {rcens}

Type:	Package
Title:	Generate Sample Censoring
Version:	0.2.2
Author:	Daniel Saavedra [aut, cre, cph], Pedro L Ramos [aut]
Maintainer:	Daniel Saavedra <dlsaavedra@uc.cl>
Description:	Provides functions to generate censored samples of type I, II and III, from any random sample generator. It also supplies the option to create left and right censorship. Along with this, the generation of samples with interval censoring is in the testing phase, with two options of fixed length intervals and random lengths. Additional functions generate complex inspection, delayed-entry, hybrid, progressive, covariate-dependent, frailty-dependent, and competing-risk observation structures while retaining a consistent and verifiable data object.
License:	MIT + file LICENSE
Encoding:	UTF-8
Suggests:	testthat (≥ 3.0.0), knitr, rmarkdown, survival
Config/testthat/edition:	3
VignetteBuilder:	knitr
URL:	https://github.com/dlsaavedra/rcens
BugReports:	https://github.com/dlsaavedra/rcens/issues
Config/roxygen2/version:	8.0.0
NeedsCompilation:	no
Packaged:	2026-07-23 05:32:17 UTC; danielsaavedramorales
Repository:	CRAN
Date/Publication:	2026-07-23 05:50:02 UTC

rcens: generators for censored samples

Description

The package contains the original basic censoring generators and the additive 'rcenscomp' family for complex observation mechanisms.

Author(s)

Maintainer: Daniel Saavedra dlsaavedra@uc.cl (ORCID) [copyright holder]

Authors:

• Daniel Saavedra dlsaavedra@uc.cl (ORCID) [copyright holder]

• Pedro L Ramos

See Also

Useful links:

• https://github.com/dlsaavedra/rcens

• Report bugs at https://github.com/dlsaavedra/rcens/issues

Generate interval censoring sample

Description

Generator of interval censored samples where the length of interval is random, given a generator of samples of the distribution X (rdistrX) with parameters appended by the list param_X. Generator sample of distribution C (censoring) with parameters appended by the list param_C In which, you can control the desired censorship percentage.

Usage

rcensI(
  rdistrX,
  rdistrC,
  param_X,
  param_C,
  n = 10000,
  epsilon = 0.5,
  n_mc = 10000,
  theta = 1,
  verbose = FALSE
)

Arguments

Value

A list with sample data information:

Author(s)

Daniel Saavedra Morales

See Also

rcensT1 for generate censorship sample type I.
rcensT2 for generate censorship sample type II.
rcensT3 for generate censorship sample type III
rcensIfix for generate interval censoring sample with fix length interval

Examples

#Example Exponential - Uniform

Data_I = rcensI(rdistrX = rexp, rdistrC = runif,
               param_X = list("rate" = 2),
              param_C = list("min" = 0, "max" = 1),
             n = 1e02, theta = .9)

## Example with plot in examples_plot/Example_rcensI_plot.R

Generate interval censoring sample (Fix)

Description

Generator of interval censored samples where the length of interval is fixed, given a generator of samples of the distribution X (rdistrX) with parameters appended by the list param_X. In which, you can control the desired censorship percentage.

Usage

rcensIfix(
  rdistrX,
  param_X,
  interval_length,
  n = 10000,
  theta = 1,
  verbose = FALSE
)

Arguments

Value

A list with sample data information:

Author(s)

Daniel Saavedra Morales

See Also

rcensT1 for generate censorship sample type I.
rcensT2 for generate censorship sample type II.
rcensT3 for generate censorship sample type III
rcensI for generate interval censoring sample with random length interval

Examples

#Example Exponential - Fix Interval

Data_Ifix = rcensIfix(rdistrX = rexp, interval_length = 2,
                   param_X = list("rate" = .5),
                   n = 1e02, theta = .9)

## Example with plot in examples_plot/Example_rcensIfix_plot.R

Generate Censoring Sample, Type I

Description

Generator of censored samples type I with right or left censoring, given a generator of samples of the distribution X (rdistrX) with parameters appended by the list param_X. In which, you can control the censorship time or the desired censorship percentage.

Usage

rcensT1(
  rdistrX,
  param_X,
  qdistrX = NULL,
  n = 10000,
  t_censored = -1,
  theta = 0.5,
  verbose = FALSE,
  right = TRUE
)

Arguments

Value

A list with sample data information:

Author(s)

Daniel Saavedra Morales

See Also

rcensT2 for generate censorship sample type II.
rcensT3 for generate censorship sample type III
rcensI for generate interval censoring sample with random length interval
rcensIfix for generate interval censoring sample with fix length interval

Examples

## Example Exponential
## time censored

Data_T1 = rcensT1(rdistrX = rexp, param_X = list("rate" = 2),
                 n = 1e02, t_censored = 1)

## time censored estimate with desired censoring percentage.

Data_T1 = rcensT1(rdistrX = rexp, param_X = list("rate" = 2),
                 qdistrX = qexp, n = 1e02, theta = .8)

## Example with plot in examples_plot/Example_rcensT1_plot.R

Generate Censoring Sample, Type II

Description

Generator of censored samples type II with right or left censoring, given a generator of samples of the distribution X (rdistrX) with parameters appended by the list param_X. In which, you can control the number of censored sample or the desired censorship percentage.

Usage

rcensT2(
  rdistrX,
  param_X,
  n = 10000,
  m_censored = -1,
  theta = 0.5,
  verbose = FALSE,
  right = TRUE
)

Arguments

Value

A list with sample data information:

Author(s)

Daniel Saavedra Morales

See Also

rcensT1 for generate censorship sample type I.
rcensT3 for generate censorship sample type III
rcensI for generate interval censoring sample with random length interval
rcensIfix for generate interval censoring sample with fix length interval

Examples

##Example Exponential

## Number of sample censored
Data_T2 = rcensT2(rdistrX = rexp, param_X = list("rate" = 2), n = 1e02, m_censored = 9)

## Number of censored sample estimate with desired censoring percentage.
Data_T2 = rcensT2(rdistrX = rexp, param_X = list("rate" = 2), n = 1e02, theta = .8)


## Example with plot in examples_plot/Example_rcensT2_plot.R

Generate Censoring Sample,Type III (Random)

Description

Generator of censored samples type III with right or left censoring, given a generator of samples of the distribution X (rdistrX) with parameters appended by the list param_X. Also accumulate function of distribution and generator sample of distribution C (censoring) with parameters appended by the list param_C. In which, you can control the desired censorship percentage.

Usage

rcensT3(
  rdistrX,
  pdistrC,
  rdistrC,
  param_X,
  param_C,
  n = 10000,
  theta = 0.5,
  n_mc = 10000,
  lambda_tol = c(1e-06, 10000),
  verbose = FALSE,
  right = TRUE
)

Arguments

Value

A list with sample data information:

Author(s)

Daniel Saavedra Morales

See Also

rcensT1 for generate censorship sample type I.
rcensT2 for generate censorship sample type II
rcensI for generate interval censoring sample with random length interval
rcensIfix for generate interval censoring sample with fix length interval

Examples

#Example Exponential - Uniform


Data_T3 = rcensT3(rdistrX = rexp, pdistrC = punif, rdistrC = runif,
              param_X = list("rate" = 2),
             param_C = list("min" = 0, "max" = "lambda"),
            n = 1e02, theta = .9, right = TRUE)


Data_T3 = rcensT3(rdistrX = rexp, pdistrC = punif, rdistrC = runif,
                 param_X = list("rate" = 2),
                 param_C = list("min" = 0, "max" = "lambda"),
                 n = 1e02, theta = .1, right = FALSE)


## Example with plot in examples_plot/Example_rcensT3_plot.R

Generate data under a complex censoring or observation scheme

Description

'rcenscomp()' is the unified entry point for the complex-censoring extension. It separates latent event generation from the observation operator and dispatches to a specialized 'rcenscomp_*()' function.

Usage

rcenscomp(
  scheme,
  x = NULL,
  ...,
  rdistrX = NULL,
  param_X = list(),
  n = NULL,
  keep_latent = TRUE
)

Arguments

Details

The supported taxonomy comprises inspection schemes ('interval', 'current_status'), delayed entry, hybrid schemes, progressive schemes, conditionally independent covariate-dependent censoring, informative censoring induced by shared process frailty, and competing risks.

For schemes based on one latent sample, provide exactly one of 'x' and 'rdistrX'. With 'rdistrX', the latent sample is generated by 'do.call(rdistrX, c(list(n), param_X))'. Schemes that generate multiple processes internally ('covariate_censoring', 'informative_frailty', and 'competing_risks') require 'n' and reject 'x' and 'rdistrX'.

Each result records the canonical scheme name and the observed data object. The likelihood appropriate to that observed object is not fitted by this package extension.

Value

An object of class 'rcenscomp' with components 'data', 'latent', 'scheme', 'design', 'diagnostics', and 'call'.

References

Ramos, P. L., Kumar, I., Dutta, S., and Kundu, D. *Sampling with Complex Censored Data: A Practical Guide* (unpublished manuscript).

See Also

• rcenscomp_interval

• rcenscomp_hybrid_type1

• rcenscomp_progressive_type2

• rcenscomp_informative_frailty

• rcenscomp_competing_risks

Examples

set.seed(2026)
dat <- rcenscomp(
  scheme = "hybrid_type1",
  rdistrX = rexp,
  param_X = list(rate = 2),
  n = 100,
  T = 1,
  r = 60
)
dat

x <- rexp(20, rate = 2)
rcenscomp("interval", x = x, visits = c(0.2, 0.5, 1))

Methods for complex censored samples

Description

'print()' gives a compact design summary, 'summary()' returns the complete design diagnostics, and 'as.data.frame()' extracts the observed data object.

Usage

## S3 method for class 'rcenscomp'
print(x, ...)

## S3 method for class 'rcenscomp'
summary(object, ...)

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

Arguments

Value

'print()' returns 'x' invisibly. 'summary()' returns an object of class 'summary.rcenscomp'. 'as.data.frame()' returns 'x$data'.

References

Ramos, P. L., Kumar, I., Dutta, S., and Kundu, D. *Sampling with Complex Censored Data: A Practical Guide* (unpublished manuscript).

Examples

x <- c(0.2, 0.5, 0.9, 1.3)
z <- rcenscomp_hybrid_type1(x, T = 1, r = 3)
print(z)
summary(z)
as.data.frame(z)

Apply adaptive progressive Type-II hybrid censoring

Description

Observes exactly 'length(R)' failures, follows planned removals through failures at or before 'T', and suppresses future intermediate removals after the ideal time is crossed.

Usage

rcenscomp_adaptive_progressive_type2(
  x,
  R,
  T,
  ties = "error",
  keep_latent = TRUE
)

Arguments

Details

Failures at times '<= T' still use the planned withdrawal. At the first pre-final failure after 'T', adaptation begins and later intermediate withdrawals are zero. All remaining units are withdrawn at the 'm'th failure. If the first crossing occurs only at the final failure, there is no future intermediate withdrawal to adapt and the 'adaptive' indicator remains false.

Value

An 'rcenscomp' object with exactly 'm' events, realized 'R_used', an adaptation indicator and, when applicable, its first index and time.

References

Ng, H. K. T., Kundu, D., and Chan, P. S. (2009). Statistical analysis under an adaptive Type-II progressive censoring scheme. *Naval Research Logistics*, 56, 687-698.

See Also

• rcenscomp_progressive_hybrid

Examples

set.seed(4)
rcenscomp_adaptive_progressive_type2(
  x = c(0.2, 0.6, 1.1, 1.5, 2),
  R = c(1, 0, 1), T = 0.5
)

Calibrate the time cap of a progressive hybrid design

Description

Chooses a time cap 'T' so the expected number of failures in a time-capped progressive hybrid design is close to 'target_D'.

Usage

rcenscomp_calibrate_progressive_hybrid_time(
  n,
  beta,
  alpha,
  R,
  target_D,
  lower,
  upper,
  B = 1000,
  grid_size = 201,
  seed = 2026
)

Arguments

Details

Latent Weibull samples and withdrawal-randomization streams are held common across every candidate 'T'. The Weibull parameterization uses shape 'alpha' and rate-like 'beta'. This is a design calibration only and does not fit a censored-data model.

Value

A list with selected 'T', achieved expected failures 'achieved_D', complete grid, seed, replication count, and requested target.

See Also

• rcenscomp_calibrate_time

• rcenscomp_progressive_hybrid

Examples

rcenscomp_calibrate_progressive_hybrid_time(
  n = 10, beta = 2.5, alpha = 1.5,
  R = c(1, 1, 0, 0, 0, 0, 0, 0),
  target_D = 5, lower = 0.1, upper = 1,
  B = 20, grid_size = 7, seed = 2026
)

Calibrate a time parameter by Monte Carlo with common random numbers

Description

Selects the grid value whose Monte Carlo design summary is closest to a requested target while reusing latent samples and randomization streams at every grid point.

Usage

rcenscomp_calibrate_time(
  generator,
  censor_fun,
  target,
  lower,
  upper,
  B = 1000,
  grid_size = 201,
  seed = 2026,
  metric = c("censoring_proportion", "expected_failures", "probability_few_failures",
    "expected_stopping_time"),
  d_min = NULL
)

Arguments

Details

'generator()' must return one latent sample. 'censor_fun(latent, value)' must return either an 'rcenscomp' object or a data frame. Supported metrics are expected censoring proportion, expected failure count, probability of fewer than 'd_min' failures, and expected stopping time. Common random numbers reduce Monte Carlo noise across grid values. The supplied 'seed' is local: the caller's '.Random.seed' is restored exactly, or removed again if it did not previously exist.

The procedure is a finite-grid calibration, not a proof that the target is attainable or that the response is monotone. A warning is emitted when the target lies outside the achieved grid range. Every replicate must produce a finite scalar summary; invalid or missing summaries are rejected explicitly.

Value

A list with selected 'value', 'achieved', complete 'grid', 'seed', 'B', 'target', 'metric', and 'd_min'.

References

Ramos, P. L., Kumar, I., Dutta, S., and Kundu, D. *Sampling with Complex Censored Data: A Practical Guide* (unpublished manuscript).

See Also

• rcenscomp_calibrate_progressive_hybrid_time

Examples

generator <- function() rexp(20, rate = 2)
censor_fun <- function(x, T) rcenscomp_hybrid_type1(x, T = T, r = 10)
rcenscomp_calibrate_time(
  generator, censor_fun, target = 0.5,
  lower = 0.1, upper = 1.5, B = 20, grid_size = 9,
  seed = 2026, metric = "censoring_proportion"
)

Generate independent-latent-time competing risks

Description

Generates one Weibull latent time for each cause, records the earliest cause, and applies optional administrative censoring.

Usage

rcenscomp_competing_risks(
  n,
  beta,
  alpha,
  censor_time = Inf,
  keep_latent = TRUE,
  ties = "error"
)

Arguments

Details

'status = 0' denotes censoring and 'status = 1, ..., K' denotes the cause of the earliest event. Independent latent cause times are a simulation assumption and are not identifiable from ordinary competing-risk data. Competing events are not ordinary independent censoring for a cumulative incidence estimade. Cause ties are rejected by default; 'ties = "first"' uses the smallest cause index as an explicit computational convention.

Value

An 'rcenscomp' object whose observed data has 'id', 'time', and integer 'status'. Diagnostics include counts for censoring and each cause.

References

Prentice, R. L. et al. (1978). The analysis of failure times in the presence of competing risks. *Biometrics*, 34, 541-554.

See Also

• rcenscomp_rweibull_rate

Examples

set.seed(22)
rcenscomp_competing_risks(
  n = 10,
  beta = c(1.2, 0.8),
  alpha = c(1.4, 1.1),
  censor_time = 1
)

Generate conditionally independent covariate-dependent censoring

Description

Generates Weibull PH event and censoring times that share recorded covariates but use independent uniforms.

Usage

rcenscomp_covariate_censoring(
  n,
  beta_x,
  alpha_x,
  gamma_x,
  beta_c,
  alpha_c,
  gamma_c,
  z = NULL,
  keep_latent = TRUE
)

Arguments

Details

Conditional on 'z', the generated event and censoring times are independent. Consequently, covariate-dependent censoring is not automatically informative when the analysis conditions on the relevant covariates. If 'z = NULL', one Bernoulli(0.5) and one standard normal covariate are generated, requiring both coefficient vectors to have length two.

Value

An 'rcenscomp' object whose observed data has 'id', 'time', 'delta', and the covariate columns.

References

Ramos, P. L., Kumar, I., Dutta, S., and Kundu, D. *Sampling with Complex Censored Data: A Practical Guide* (unpublished manuscript).

See Also

• rcenscomp_rweibull_ph

• rcenscomp_informative_frailty

Examples

set.seed(20)
rcenscomp_covariate_censoring(
  n = 10,
  beta_x = 2.5, alpha_x = 1.5, gamma_x = c(0.5, -0.3),
  beta_c = 1.2, alpha_c = 1.1, gamma_c = c(-0.2, 0.4)
)

Generate current-status observations

Description

Records only whether each latent event occurred by its single inspection time.

Usage

rcenscomp_current_status(x, inspection, keep_latent = TRUE)

Arguments

Details

The observed status is 'as.integer(x <= inspection)'. When it equals one, 'inspection' is not the exact failure time; the event is known only to have occurred in '(0, inspection]'. The function assumes the inspection process is supplied externally and does not assert ignorability on its own.

Value

An 'rcenscomp' object whose observed 'data' has 'id', 'inspection', and binary 'status' columns. Diagnostics give the number and proportion of positive current-status records.

References

Groeneboom, P. and Wellner, J. A. (1992). *Information Bounds and Nonparametric Maximum Likelihood Estimation*. Birkhauser.

See Also

• rcenscomp_interval

Examples

rcenscomp_current_status(
  x = c(0.2, 0.8, 1.5),
  inspection = c(0.3, 0.5, 2)
)

Apply delayed entry and post-entry right censoring

Description

Excludes units with 'x <= entry', then records right-censored follow-up for the units that survived long enough to enter observation.

Usage

rcenscomp_delayed_entry(x, entry, censor, keep_latent = TRUE)

Arguments

Details

Units excluded by 'x <= entry' are left truncated, not censored. Their IDs remain absent from the observed table. For included units, 'time' is 'pmin(x, censor)' and 'delta' is 'as.integer(x <= censor)'. Valid analysis must condition on survival to entry; this generator does not fit that likelihood. It assumes the supplied entry and post-entry censoring process is appropriate for the intended simulation.

Value

An 'rcenscomp' object whose observed 'data' has 'id', 'entry', 'time', and 'delta'. Diagnostics distinguish truncation from post-entry censoring and give total post-entry time at risk.

References

Andersen, P. K., Borgan, O., Gill, R. D., and Keiding, N. (1993). *Statistical Models Based on Counting Processes*. Springer.

See Also

• rcenscomp_current_status

Examples

rcenscomp_delayed_entry(
  x = c(0.2, 0.8, 1.5, 2),
  entry = c(0.3, 0.1, 0.5, 1),
  censor = c(1, 1, 1.2, 3)
)

Apply generalized hybrid censoring

Description

Uses two failure thresholds and an intermediate planned time to bound the realized number of failures under the continuous-time convention.

Usage

rcenscomp_generalized_hybrid(x, T, r1, r2, keep_latent = TRUE)

Arguments

Details

The stopping time is exactly 'min(max(T, sort(x)[r1]), sort(x)[r2])'. With continuous lifetimes, 'r1 <= D <= r2'. Ties at the stopping time are not broken silently and produce a warning because they can violate the nominal upper count.

Value

An 'rcenscomp' object with subject-wise right-censored data and design fields 'Tstar', 'D', 'r1', and 'r2'.

References

Chandrasekar, B., Childs, A., and Balakrishnan, N. (2004). Exact likelihood inference under generalized hybrid censoring. *Naval Research Logistics*, 51, 994-1004.

See Also

• rcenscomp_hybrid_type1

• rcenscomp_hybrid_type2

Examples

rcenscomp_generalized_hybrid(
  c(0.2, 0.5, 0.9, 1.4, 2), T = 0.8, r1 = 2, r2 = 4
)

Apply Hybrid Type-I censoring

Description

Stops a complete life test at the earlier of a time limit and the 'r'th ordered latent failure.

Usage

rcenscomp_hybrid_type1(x, T, r, keep_latent = TRUE)

Arguments

Details

The stopping time is 'Tstar = min(T, sort(x)[r])'. Subject-wise observations are 'time = pmin(x, Tstar)' and 'delta = as.integer(x <= Tstar)'. The number of failures may be smaller than 'r'. Ties at the stopping time generate a warning because they can change the nominal event count; the design is theoretically stated for continuous lifetimes.

Value

An 'rcenscomp' object with subject-wise 'id', 'time', and 'delta', plus 'Tstar' and the realized failure count 'D'.

References

Childs, A., Chandrasekar, B., Balakrishnan, N., and Kundu, D. (2003). Exact likelihood inference based on Type-I and Type-II hybrid censored samples. *Annals of the Institute of Statistical Mathematics*, 55, 319-330.

See Also

• rcenscomp_hybrid_type2

• rcenscomp_generalized_hybrid

Examples

rcenscomp_hybrid_type1(c(0.2, 0.5, 0.9, 1.4), T = 0.8, r = 3)

Apply Hybrid Type-II censoring

Description

Stops a complete life test at the later of a time threshold and the 'r'th ordered latent failure.

Usage

rcenscomp_hybrid_type2(x, T, r, keep_latent = TRUE)

Arguments

Details

The stopping time is 'Tstar = max(T, sort(x)[r])'. Under continuous event times, at least 'r' failures are observed. Ties at the stopping point are reported because they may produce more events than a nominal target.

Value

An 'rcenscomp' object with subject-wise 'id', 'time', and 'delta', plus 'Tstar' and the realized failure count 'D'.

References

Balakrishnan, N. and Kundu, D. (2013). Hybrid censoring: models, inferential results and applications. *Computational Statistics & Data Analysis*, 57, 166-209.

See Also

• rcenscomp_hybrid_type1

• rcenscomp_generalized_hybrid

Examples

rcenscomp_hybrid_type2(c(0.2, 0.5, 0.9, 1.4), T = 0.8, r = 3)

Generate informative censoring through shared process frailty

Description

Generates event and censoring times that share an unobserved individual gamma frailty, with exponent 'rho' in the censoring process.

Usage

rcenscomp_informative_frailty(
  n,
  beta_x,
  alpha_x,
  beta_c,
  alpha_c,
  gamma_x,
  gamma_c,
  theta,
  rho,
  z = NULL,
  keep_latent = TRUE
)

Arguments

Details

Frailty is 'Gamma(shape = 1 / theta, scale = theta)'. The event linear predictor adds 'log(V_i)' and the censoring predictor adds 'rho * log(V_i)'. The two processes use separate uniforms conditional on frailty and covariates. This is a sensitivity generator, not an inferential correction, and the full joint mechanism is generally not identified by the observed '(time, delta, z)' table alone. The boundary 'rho = 0' removes frailty from the censoring predictor, making censoring independent of the event time conditional on 'z'. At that boundary the censoring mechanism is ignorable, but this does not imply that an ordinary event model correctly handles omitted event frailty.

Value

An 'rcenscomp' object with observed 'id', 'time', 'delta', and covariates, plus separate latent process values when requested.

References

Lagakos, S. W. (1979). General right censoring and its impact on survival analysis. *Biometrics*, 35, 139-156.

See Also

• rcenscomp_covariate_censoring

• rcenscomp_rweibull_ph_frailty

Examples

set.seed(21)
rcenscomp_informative_frailty(
  n = 10,
  beta_x = 2.5, alpha_x = 1.5,
  beta_c = 1.3, alpha_c = 1.1,
  gamma_x = c(0.5, -0.3), gamma_c = c(-0.2, 0.4),
  theta = 0.6, rho = 0.8
)

Apply a common visit schedule to latent event times

Description

Converts exact latent event times into left-, interval-, or right-censored records under a common inspection schedule.

Usage

rcenscomp_interval(x, visits, keep_latent = TRUE)

Arguments

Details

For visits '0 < v1 < ... < vK', an event at or before 'v1' is recorded as '(0, v1]'; an event after 'vK' is recorded as '(vK, Inf]'; all other events are assigned to the unique interval '(v[j-1], v[j]]'. This differs from the original [rcensI()] and [rcensIfix()] functions: those functions retain their own censoring-control strategies, while this function applies a fixed inspection operator to every supplied latent lifetime.

Value

An 'rcenscomp' object whose observed 'data' has columns 'id', 'L', 'R', and 'status'. Diagnostics include counts by censoring type and the finite interval widths.

References

Turnbull, B. W. (1976). The empirical distribution function with arbitrarily grouped, censored and truncated data. *JRSS B*, 38, 290-295.

Ramos, P. L., Kumar, I., Dutta, S., and Kundu, D. *Sampling with Complex Censored Data: A Practical Guide* (unpublished manuscript).

See Also

• rcenscomp_current_status

• rcensI

• rcensIfix

Examples

rcenscomp_interval(
  x = c(0.1, 0.2, 0.35, 0.6, 1.2),
  visits = c(0.2, 0.5, 1)
)

Apply time-capped progressive hybrid censoring

Description

Follows a progressive failure-time withdrawal plan until the target failure count is reached or the next failure would occur after a time cap.

Usage

rcenscomp_progressive_hybrid(x, R, T, ties = "error", keep_latent = TRUE)

Arguments

Details

'm' is 'length(R)'. If the next eligible failure exceeds 'T', all survivors are censored at 'T'. If the 'm'th failure is reached first, all remaining survivors are censored at that failure time. The returned 'R_used' contains realized withdrawals after failures; 'Rstar' is the terminal censoring count. 'D = 0' is valid and is reported diagnostically rather than treated as an estimation-ready sample.

Value

An 'rcenscomp' object with subject-wise data and design fields 'D', 'Tstar', 'R_used', 'Rstar', and the original 'R'.

References

Kundu, D. and Joarder, A. (2006). Analysis of Type-II progressively hybrid censored data. *Computational Statistics & Data Analysis*, 50, 2509-2528.

See Also

• rcenscomp_progressive_type2

• rcenscomp_adaptive_progressive_type2

Examples

set.seed(3)
rcenscomp_progressive_hybrid(
  x = c(0.2, 0.5, 0.9, 1.4, 2),
  R = c(1, 0, 1), T = 0.8
)
rcenscomp_progressive_hybrid(
  x = c(1, 2, 3), R = c(0, 0, 0), T = 0.5
)

Apply progressive Type-I censoring

Description

Continuously observes failures while withdrawing survivors at fixed times and censoring every remaining unit at the final scheduled time.

Usage

rcenscomp_progressive_type1(
  x,
  times,
  R_plan = integer(0),
  ties = "error",
  keep_latent = TRUE
)

Arguments

Details

Before each scheduled withdrawal, every still-alive unit with an event time at or before that schedule time is recorded at its exact event time. At an intermediate time, the realized withdrawal is the smaller of the planned count and the eligible risk set. At the final time, all survivors are censored. This is not progressively interval-censored data: failures are monitored continuously.

Value

An 'rcenscomp' object with subject-wise right-censored data and realized withdrawals 'R_used', including the terminal withdrawal.

References

Balakrishnan, N. and Cramer, E. (2014). *The Art of Progressive Censoring*. Birkhauser.

See Also

• rcenscomp_progressive_type2

Examples

set.seed(2)
rcenscomp_progressive_type1(
  x = c(0.1, 0.3, 0.7, 1.2, 2),
  times = c(0.25, 0.75, 1.5),
  R_plan = c(1, 0)
)

Apply progressive Type-II censoring

Description

Observes exactly 'length(R)' failures and withdraws the planned numbers of eligible survivors at their corresponding failure times.

Usage

rcenscomp_progressive_type2(x, R, ties = "error", keep_latent = TRUE)

Arguments

Details

At each observed failure, the failed unit is removed before 'R[j]' survivors are sampled uniformly without replacement. Validity requires non-negative integer withdrawals with 'sum(R) = n - length(R)'. The default rejects tied latent times. With 'ties = "first"', the smallest original ID is the deterministic failure among a tied risk set; this is a computational convention, not part of the continuous-time design.

Value

An 'rcenscomp' object with subject-wise 'id', 'time', and 'delta', and a compact table containing 'failure_order', 'failure_id', 'failure_time', and 'R_used'.

References

Balakrishnan, N. and Aggarwala, R. (2000). *Progressive Censoring: Theory, Methods, and Applications*. Birkhauser.

See Also

• rcenscomp_progressive_type1

• rcenscomp_progressive_hybrid

Examples

set.seed(1)
rcenscomp_progressive_type2(
  x = c(0.2, 0.4, 0.8, 1.1, 1.5),
  R = c(1, 0, 1)
)

Generate covariate-dependent Weibull censoring times

Description

Generates censoring times from their own Weibull proportional-hazards model.

Usage

rcenscomp_rweibull_censor_ph(z, beta_c, alpha_c, gamma_c)

Arguments

Details

Each call consumes a new independent set of uniforms. Independence from an event generator therefore follows when the caller supplies a common 'z' and invokes the two generators sequentially without reusing random draws.

Value

A numeric vector of censoring times.

See Also

• rcenscomp_rweibull_ph

• rcenscomp_covariate_censoring

Examples

set.seed(12)
z <- cbind(z1 = c(0, 1, 0), z2 = c(-1, 0, 1))
rcenscomp_rweibull_censor_ph(z, 1.2, 1.1, c(-0.2, 0.4))

Generate Weibull proportional-hazards lifetimes

Description

Generates Weibull event times with cumulative hazard 'beta * t^alpha * exp(z

Usage

rcenscomp_rweibull_ph(n, beta, alpha, gamma, z = NULL)

Arguments

Details

The event time is '(-log(U) / (beta * exp(z_i^T gamma)))^(1 / alpha)'. If 'z = NULL', the documented didactic default generates one Bernoulli(0.5) and one standard normal covariate, so 'gamma' must then have length two. A supplied 'z' may have any number of numeric columns compatible with 'gamma'. No intercept is added and no seed is set.

Value

A data frame containing 'id', latent event time 'x', and all covariates.

References

Bender, R., Augustin, T., and Blettner, M. (2005). Generating survival times to simulate Cox proportional hazards models. *Statistics in Medicine*, 24, 1713-1723.

See Also

• rcenscomp_rweibull_rate

• rcenscomp_rweibull_censor_ph

• rcenscomp_rweibull_ph_frailty

Examples

set.seed(11)
rcenscomp_rweibull_ph(
  5, beta = 2.5, alpha = 1.5, gamma = c(0.5, -0.3)
)

Generate clustered Weibull PH lifetimes with shared gamma frailty

Description

Generates a latent event-time model in which all units in a cluster share one gamma frailty. This is an event generator, not a censoring mechanism.

Usage

rcenscomp_rweibull_ph_frailty(
  cluster_size,
  beta,
  alpha,
  gamma,
  theta,
  z = NULL
)

Arguments

Details

Frailties use 'shape = 1 / theta' and 'scale = theta', giving mean one and variance 'theta'. The event linear predictor adds 'log(V_cluster)'. If 'z = NULL', the same two-covariate default as [rcenscomp_rweibull_ph()] is used. The marginal event hazard after integrating frailty is not the baseline conditional hazard.

Value

A data frame with 'id', 'cluster', latent 'x', repeated cluster 'frailty', and covariates.

References

Hougaard, P. (2000). *Analysis of Multivariate Survival Data*. Springer.

See Also

• rcenscomp_rweibull_ph

• rcenscomp_informative_frailty

Examples

set.seed(13)
rcenscomp_rweibull_ph_frailty(
  cluster_size = c(2, 3), beta = 2.5, alpha = 1.5,
  gamma = c(0.5, -0.3), theta = 0.6
)

Generate Weibull lifetimes under the manuscript rate parameterization

Description

Generates independent Weibull event times with density 'alpha * beta * x^(alpha - 1) * exp(-beta * x^alpha)'.

Usage

rcenscomp_rweibull_rate(n, beta, alpha)

Arguments

Details

The inverse transform is '(-log(U) / beta)^(1 / alpha)'. Here 'alpha' is shape and 'beta' is rate-like. The equivalent [stats::rweibull()] scale is 'beta^(-1 / alpha)'. A log-scale equivalent is used when the direct calculation has an invalid intermediate result but the lifetime remains representable as a positive finite double. Unrepresentable results raise an error. No seed is set internally.

Value

A numeric vector of positive latent lifetimes.

References

Weibull, W. (1951). A statistical distribution function of wide applicability. *Journal of Applied Mechanics*, 18, 293-297.

See Also

• rweibull

• rcenscomp_rweibull_ph

Examples

set.seed(10)
rcenscomp_rweibull_rate(5, beta = 2.5, alpha = 1.5)

Title Generate Sample with Cure Fraction

Description

Generator Sample with Cure Fraction, given a generator of samples of the distribution X (rdistrX) with parameters appended by the list param_X. Also the proportion of cure desired p.

Usage

rcuref(rdistrX, param_X, n = 10000, p = 0.5)

Arguments

Value

A list with sample data information:

Author(s)

Daniel Saavedra Morales

Examples

#Example Exponential Cure Fraction p = 0.5

Data = rcuref(rdistrX = rexp, param_X = list("rate" = 1),
              n = 1000, p = 0.5)

Title Generate Sample with Cure Fraction and Random Censoring

Description

Generator Sample with Cure Fraction, Random Censoring. Given a generator of samples of the distribution X (rdistrX) with parameters appended by the list param_X. Also accumulate function of distribution and generator sample of distribution C (censoring) with parameters appended by the list param_C In which, you can control the desired censorship percentage.
Note: cure fraction (p) must be less than desired censorship percentage.

Usage

rcurefT3(
  rdistrX,
  pdistrC,
  rdistrC,
  param_X,
  param_C,
  p = 0.1,
  n = 10000,
  theta = 0.5,
  n_mc = 10000,
  lambda_tol = c(1e-06, 10000),
  verbose = FALSE,
  right = TRUE
)

Arguments

Value

A list with sample data information:

Author(s)

Daniel Saavedra Morales

See Also

rcuref Generate Sample with Cure Fraction.

Examples

#Example Exponential - Uniform


Data_T3 = rcurefT3(rdistrX = rexp, pdistrC = punif, rdistrC = runif,
                   param_X = list("rate" = 2),
                   param_C = list("min" = 0, "max" = "lambda"),
                   n = 1e02, theta = .9, p = 0.2)




## Example with plot in examples_plot/Example_rcurefT3_plot.R