Package {ZTRecur}

Type: Package
Title: Analysis for Zero-Truncated Recurrent Event Data
Version: 0.2.1
Date: 2026-10-17
Description: Provides nonparametric and Cox-type regression methods for analyzing zero-truncated recurrent event data with supplementary information. Key features include estimating rate and mean functions under nonparametric marginal models; estimating time-constant regression coefficients under standard and stratified Andersen-Gill (AG) models; and estimating time-varying covariate effects under stratified and nonstratified Cox-type models. Standard errors (SEs) are estimated using sandwich-based variance estimators for nonstratified models and resampling-based procedures for stratified models. For further details, see Chen, Hu, and Rosychuk (2026) <doi:10.1214/25-AOAS2130>, Chen et al. (2026) <doi:10.1002/cjs.70064>, and Chen, Hu, and Rosychuk (2026) <doi:10.48550/arXiv.2607.26210>.
License: GPL-2 | GPL-3 [expanded from: GPL (≥ 2)]
Encoding: UTF-8
Imports: Rcpp (≥ 1.0.0), dplyr, graphics, grDevices, stats
LinkingTo: Rcpp, RcppArmadillo
NeedsCompilation: yes
ByteCompile: yes
Depends: R (≥ 3.5)
LazyData: true
Config/roxygen2/version: 8.0.0
Suggests: testthat (≥ 3.0.0)
Config/testthat/edition: 3
Packaged: 2026-10-27 10:13:53 UTC; AnqiChen
Author: Anqi A. Chen [aut, cre], X. Joan Hu [aut], Yi Xiong [aut], Rhonda J. Rosychuk [aut]
Maintainer: Anqi A. Chen <anqi6@ualberta.ca>
Repository: CRAN
Date/Publication: 2026-11-12 11:10:10 UTC

ZTRecur: Analysis for Zero-Truncated Recurrent Event Data

Description

This package provides nonparametric and Cox-type regression methods for analyzing zero-truncated recurrent event data with supplementary information. Key features include estimating rate and mean functions under nonparametric marginal models; estimating time-constant regression coefficients under standard and stratified Andersen-Gill (AG) models; and estimating time-varying covariate effects under stratified and nonstratified Cox-type models. Standard errors (SEs) are estimated using sandwich-based variance estimators for nonstratified models and resampling-based procedures for stratified models. For further details, see Chen, Hu, and Rosychuk (2026) doi:10.1214/25-AOAS2130, Chen et al. (2026) doi:10.1002/cjs.70064, and Chen, Hu, and Rosychuk (2026) https://arxiv.org/abs/2607.26210.

Author(s)

Maintainer: Anqi A. Chen anqi6@ualberta.ca

Authors:

Generate Birthdate Intervals Based on Integer Ages at Event Times

Description

Generate Birthdate Intervals Based on Integer Ages at Event Times

Usage

birth.interval(visit, MAXAGE = 18, WINDOWL, WINDOWR)

Arguments

visit

Matrix of event-level information. It contains subject ID (patid), Julian date of the event (date), and integer age at that event date (age).

MAXAGE

Maximum age considered. The default value is 18.

WINDOWL

The start date of data extraction window, such as as.Date("2010-04-01").

WINDOWR

The end date of data extraction window, such as as.Date("2025-03-31").

Details

The calculation of birthdate interval is described in Hu & Rosychuk (2016).

Value

A subject-level data frame with patient ID (patid), left (birthL) and right (birthR) endpoints of birthdate intervals in Julian date.

References

Hu, X. J., & Rosychuk, R. J. (2016). Marginal regression analysis of recurrent events with coarsened censoring times. Biometrics, 72(4), 1113-1122.

Examples

# Note: The \code{visit} must be a data frame.
birth.b <- birth.interval(visit=as.data.frame(my.visit),
WINDOWL=as.Date("2010-04-01"), WINDOWR=as.Date("2025-03-31"))

Sample Population Aggregate Dataset

Description

A sample data frame containing aggregated population census information.

Usage

my.agge.info

Format

A data frame with $n$ rows and the following variables:

AGE

Integer age.

MaleInd

Indicator for male sex.

CityAInd

Indicator for residence in region A.

CityBInd

Indicator for residence in region B.

pop

Number of subjects in the corresponding demographic group.

Sample Base Dataset

Description

A sample matrix containing subject-level demographic information and the number of observed events for each subject.

Usage

my.base

Format

A matrix with $n$ rows and the following variables:

patid

Subject ID.

MaleInd

Indicator for male sex.

CityAInd

Indicator for residence in region A.

CityBInd

Indicator for residence in region B.

nmvst

Number of observed events.

p1stvt

Row index of the subject's first observed event in the my.visit matrix.

birthl

Lower endpoint of the birthdate interval, represented as a Julian date.

birthr

Upper endpoint of the birthdate interval, represented as a Julian date.

birthJulian

Exact birthdate, represented as a Julian date.

Sample Visit Dataset

Description

A sample matrix containing event-level information for each subject.

Usage

my.visit

Format

A matrix with $n$ rows and the following variables:

patid

Subject ID.

reptind

Indicator for a subsequent event: 0 for the first observed event and 1 for subsequent events for the subject.

date

Julian date of the event.

age

Integer age at the date of the event.

Nonparametric Estimation of Marginal Mean and Rate Functions

Description

Performs nonparametric estimation of marginal mean and rate functions across covariate groups using zero-truncated recurrent event data integrated with supplementary population information. The procedure accounts for coarsened individual follow-up information and provides variance estimates via bootstrap or multiplier resampling techniques.

Usage

nonpara_esti(
  base,
  visit,
  mycovars,
  AGEUNIT,
  ONUM,
  WINDOWL,
  WINDOWR,
  agge.info,
  knownBirth = FALSE,
  reptw = 100,
  B = 200,
  seeds = 1234,
  resampling = FALSE,
  r.distn = "std Normal",
  save.res = FALSE,
  resultpath = NULL
)

Arguments

base

Matrix of subject-level information. Each row corresponds to one subject and contains demographic information (covariates and either the exact birthdate (birthJulian) or birthdate interval (birthl and birthr)), the number of observed events (nmvst), and the row index of the subject's first observed event in the visit matrix (p1stvt). The variables must be provided in the following order: covariates, nmvst, p1stvt, and either birthJulian or birthl and birthr.

visit

Matrix of event-level information. Each row corresponds to one observed event and contains the subject ID (patid), an indicator of whether the event is a subsequent visit (reptind; 0 for the first observed visit and 1 for subsequent visits for the same subject), the Julian date of the event (date), and the integer age at the time of the event (age). Subject IDs must be integers from 1 through the number of rows in the base matrix. If birthdate information is unavailable, the integer age (age) must be provided for each event.

mycovars

Character vector specifying the variable names of covariates to define subgroups.

AGEUNIT

Integer; the number of time units per year (e.g., 6 for bimonthly data).

ONUM

Integer; the maximum number of time units under observation (e.g., 18 * AGEUNIT for <18 years).

WINDOWL

Numeric; Julian date representing the left boundary of the data extraction window.

WINDOWR

Numeric; Julian date representing the right boundary of the data extraction window.

agge.info

Data frame containing aggregated population census data, including integer age (AGE),mycovars, and number of subjects in that subgroup (pop).

knownBirth

Logical; indicates whether exact birthdates are known. Defaults to FALSE. If FALSE, exact birthdates are unknown, and birth.interval can be used to calculate the corresponding birthdate intervals in Julian dates.

reptw

Integer; number of birthdate sets generated for the missing birthdate procedure A (Hu & Rosychuk, 2016). Default is 100.

B

Integer; number of bootstrap samples or resampling replicates for variance estimation. Default is 200.

seeds

Numeric; seed value for reproducible sampling. Default is 1234.

resampling

Logical; if TRUE, applies the multiplier resampling method described in Chen, Hu, & Rosychuk (2026). If FALSE (default), standard bootstrapping is used.

r.distn

Character string specifying the multiplier distribution used for resampling when resampling = TRUE. Options are the standard normal distribution ("std Normal"; default) or a Poisson distribution with variance 1 ("Poisson").

save.res

Logical; if TRUE, writes output files to disk. Default is FALSE.

resultpath

Character string indicating the file path destination for saved results, such as "C:\project\". Required when save.res = TRUE.

Details

The detailed estimation procedure is described in Chen, Hu & Rosychuk (2026a). The approach accounts for missing birthdates using Procedure A of Hu & Rosychuk (2016) and calculates variance estimates using the multiplier resampling procedure described in Chen, Hu & Rosychuk (2026b).

Value

A list containing the estimated marginal mean functions and rate functions, their corresponding variance estimates, and numerical results for each bootstrap or resampling sample.

References

Chen, A. A., Hu, X. J., & Rosychuk, R. J. (2026a). Statistical Learning of Pediatric Mental Health-Related Emergency Department Visits Across COVID-19 Pandemic Periods. arXiv preprint arXiv:2607.26210.

Chen, A. A., Hu, X. J., & Rosychuk, R. J. (2026b). Stratified regression analysis of zero-truncated recurrent event data. The Annals of Applied Statistics, 20(1), 434-451.

Hu, X. J., & Rosychuk, R. J. (2016). Imputation procedures for missing birthdates in administrative health data. Statistics in Medicine, 35(14), 2341-2355.

Examples


# When birthdates are unknown, bootstrap-based variance estimates are used.
nonpara.res.uk <- nonpara_esti(base = my.base[,c(-1,-9)], 
visit = my.visit, mycovars = c("MaleInd","CityAInd","CityBInd"),
  AGEUNIT = 6, ONUM = (18*6), WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), agge.info = my.agge.info,
  seeds = 1234)
  
# When birthdates are known, bootstrap-based variance estimates are used.
nonpara.res.k <- nonpara_esti(base = my.base[,c(-1,-7,-8)], 
visit = my.visit, mycovars = c("MaleInd","CityAInd","CityBInd"),
  AGEUNIT = 6, ONUM = (18*6), WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), agge.info = my.agge.info,
  knownBirth=TRUE,seeds = 1234) 
  
# When birthdates are known, multiplier resampling-based variance estimates
# are obtained using standard normal multipliers.
nonpara.res.k.r.std <- nonpara_esti(base = my.base[,c(-1,-7,-8)], 
visit = my.visit, mycovars = c("MaleInd","CityAInd","CityBInd"),
  AGEUNIT = 6, ONUM = (18*6), WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), agge.info = my.agge.info,
  knownBirth=TRUE,seeds = 1234,resampling=TRUE) 
  
# When birthdates are known, multiplier resampling-based variance estimates
# are obtained using Poisson multipliers.
nonpara.res.k.r.poi <- nonpara_esti(base = my.base[,c(-1,-7,-8)], 
visit = my.visit, mycovars = c("MaleInd","CityAInd","CityBInd"),
  AGEUNIT = 6, ONUM = (18*6), WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), agge.info = my.agge.info,
  knownBirth=TRUE,seeds = 1234,resampling=TRUE,r.distn="Poisson")

Non-Stratified Regression Analysis for Zero-Truncated Recurrent Event Data

Description

Fits a non-stratified Cox-type model with time-constant or time-varying regression coefficients using zero-truncated recurrent event data integrated with population census information. The method accounts for coarsened follow-up information and yields sandwich-form variance estimates along with empirical bootstrap or multiplier resampling alternatives.

Usage

nonstrati.reg(
  base,
  visit,
  AGEUNIT,
  ONUM,
  WINDOWL,
  WINDOWR,
  etapara,
  reptmax,
  windowKH,
  tauL,
  tauR,
  method,
  N_star,
  n,
  knownBirth = FALSE,
  reptw = 100,
  B = NA,
  seeds = 1234,
  resampling = FALSE,
  r.distn = "std Normal",
  mycovars,
  save.res = FALSE,
  resultpath = NULL,
  output.lambda = FALSE
)

Arguments

base

Matrix of subject-level information. Each row corresponds to one subject and contains demographic information (covariates and either the exact birthdate (birthJulian) or birthdate interval (birthl and birthr)), the number of observed events (nmvst), and the row index of the subject's first observed event in the visit matrix (p1stvt). The variables must be provided in the following order: covariates, nmvst, p1stvt, and either birthJulian or birthl and birthr.

visit

Matrix of event-level information. Each row corresponds to one observed event and contains the subject ID (patid), an indicator of whether the event is a subsequent visit (reptind; 0 for the first observed visit and 1 for subsequent visits for the same subject), the Julian date of the event (date), and the integer age at the time of the event (age). Subject IDs must be integers from 1 through the number of rows in the base matrix. If birthdate information is unavailable, the integer age (age) must be provided for each event.

AGEUNIT

Integer; the number of time units per year (e.g., 6 for bimonthly data).

ONUM

Integer; the maximum number of time units under observation (e.g., 18 * AGEUNIT for <18 years).

WINDOWL

Numeric; Julian date representing the left boundary of the data extraction window.

WINDOWR

Numeric; Julian date representing the right boundary of the data extraction window.

etapara

Numeric; tolerance parameter controlling the convergence criterion when solving the estimating equations.

reptmax

Integer; the maximum number of iterations allowed when solving estimating equations.

windowKH

Numeric; bandwidth size of the smoothing kernel, specified in time units.

tauL

Numeric (in time units); left boundary used for local constant estimation to avoid boundary issues.

tauR

Numeric (in time units); right boundary used for local constant estimation to avoid boundary issues.

method

Character string specifying the estimation method. Options are estimation under the Andersen-Gill model ("L"), local constant estimation under a Cox-type model with time-varying coefficients ("LC"), or both estimation procedures ("all").

N_star

Data frame containing aggregated population census data, including integer age (AGE),mycovars, and number of subjects in that subgroup (pop).

n

Integer; the approximate size of $O$, the random sample drawn from the target population.

knownBirth

Logical; indicates whether exact birthdates are known. Defaults to FALSE. If FALSE, exact birthdates are unknown, and birth.interval can be used to calculate the corresponding birthdate intervals in Julian dates.

reptw

Integer; number of birthdate sets generated for the missing birthdate procedure A (Hu & Rosychuk, 2016). Default is 100.

B

Integer or NA; number of bootstrap or resampling replicates for variance estimation. Set to NA (default) to compute only the robust sandwich-form variance estimator.

seeds

Numeric; seed value for reproducible sampling. Default is 1234.

resampling

Logical; if TRUE, applies the multiplier resampling method described in Chen, Hu, & Rosychuk (2026). If FALSE (default), standard bootstrapping is used.

r.distn

Character string specifying the multiplier distribution used for resampling when resampling = TRUE. Options are the standard normal distribution ("std Normal"; default) or a Poisson distribution with variance 1 ("Poisson").

mycovars

Character vector specifying the variable names of covariates to define subgroups.

save.res

Logical; if TRUE, writes output files to disk. Default is FALSE.

resultpath

Character string indicating the file path destination for saved results, such as "C:\project\". Required when save.res = TRUE.

output.lambda

Logical; if TRUE, includes baseline function estimates in the returned output. Default is FALSE.

Details

The estimation procedure is similar to that described in Xiong, Hu & Rosychuk (2026). The approach accounts for missing birthdates using Procedure A of Hu & Rosychuk (2016) and obtains variance estimates using the multiplier resampling procedure described in Chen, Hu & Rosychuk (2026).

Value

A list containing estimates of the regression coefficients, standard error estimates (sandwich and/or bootstrap- or resampling-based), estimated cumulative baseline functions, and estimated baseline functions when output.lambda = TRUE.

References

Chen, A. A., Hu, X. J., & Rosychuk, R. J. (2026). Stratified regression analysis of zero-truncated recurrent event data. The Annals of Applied Statistics, 20(1), 434-451.

Hu, X. J., & Rosychuk, R. J. (2016). Imputation procedures for missing birthdates in administrative health data. Statistics in Medicine, 35(14), 2341-2355.

Xiong, Y., Hu, X. J., & Rosychuk, R. J. (2026). Learning differences between two decades of mental health related emergency department visits by youth via recurrent events data analyses. Journal of the Royal Statistical Society Series C: Applied Statistics, qlag001.

Examples


# Example non-stratified regression model with unknown birthdates.
# Sandwich-form variance estimates.
nonstrati.fit.uk <- nonstrati.reg(base = my.base[,c(-1,-9)], visit = my.visit, 
AGEUNIT = 6, ONUM = 108, 
  WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), 
  etapara = 1e-4, reptmax = 50, windowKH = rep(9,108), 
  tauL = 9, tauR = 105, method = "all", N_star = my.agge.info, n=50000, 
  knownBirth = FALSE, reptw = 100, B = NA, seeds = 12345, resampling = TRUE, 
  r.distn = "std Normal",mycovars = c("MaleInd","CityAInd","CityBInd"), 
  save.res = FALSE, output.lambda = TRUE)
  
  
# Example non-stratified regression model with known birthdates.
# Sandwich-form variance estimates.
nonstrati.fit.k <- nonstrati.reg(base = my.base[,c(-1,-7,-8)], 
visit = my.visit, 
AGEUNIT = 6, ONUM = 108, 
  WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), 
  etapara = 1e-4, reptmax = 50, windowKH = rep(9,108), 
  tauL = 9, tauR = 105, method = "all", N_star = my.agge.info, n=50000, 
  knownBirth = TRUE, reptw = 100, B = NA, seeds = 12345, resampling = TRUE, 
  r.distn = "std Normal",mycovars = c("MaleInd","CityAInd","CityBInd"), 
  save.res = FALSE, output.lambda = TRUE)
  
# Example with known birthdates and bootstrap-based variance estimates.
nonstrati.fit.k.B <- nonstrati.reg(base = my.base[,c(-1,-7,-8)], 
visit = my.visit, 
AGEUNIT = 6, ONUM = 108, 
  WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), 
  etapara = 1e-4, reptmax = 50, windowKH = rep(9,108), 
  tauL = 9, tauR = 105, method = "all", N_star = my.agge.info, n=50000, 
  knownBirth = TRUE, reptw = 100, B = 200, seeds = 12345, resampling = FALSE, 
  r.distn = "std Normal",mycovars = c("MaleInd","CityAInd","CityBInd"), 
  save.res = FALSE, output.lambda = TRUE)
  
# Example with known birthdates and multiplier resampling-based variance
# estimates using standard normal multipliers.
nonstrati.fit.k.R <- nonstrati.reg(base = my.base[,c(-1,-7,-8)], 
visit = my.visit, 
AGEUNIT = 6, ONUM = 108, 
  WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), 
  etapara = 1e-4, reptmax = 50, windowKH = rep(9,108), 
  tauL = 9, tauR = 105, method = "all", N_star = my.agge.info, n=50000, 
  knownBirth = TRUE, reptw = 100, B = 200, seeds = 12345, resampling = TRUE, 
  r.distn = "std Normal",mycovars = c("MaleInd","CityAInd","CityBInd"), 
  save.res = FALSE, output.lambda = TRUE)
  
# Example with known birthdates and multiplier resampling-based variance
# estimates using Poisson multipliers with variance 1.
nonstrati.fit.k.R.p <- nonstrati.reg(base = my.base[,c(-1,-7,-8)], 
visit = my.visit, 
AGEUNIT = 6, ONUM = 108, 
  WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), 
  etapara = 1e-4, reptmax = 50, windowKH = rep(9,108), 
  tauL = 9, tauR = 105, method = "all", N_star = my.agge.info, n=50000, 
  knownBirth = TRUE, reptw = 100, B = 200, seeds = 12345, resampling = TRUE, 
  r.distn = "Poisson",mycovars = c("MaleInd","CityAInd","CityBInd"), 
  save.res = FALSE, output.lambda = TRUE)

Plot Nonparametric Marginal Mean or Rate Estimates

Description

Displays estimated marginal mean or rate functions along with pointwise confidence intervals for different covariate groups based on output from nonpara_esti.

Usage

plot_nonpar(
  res.nonpar,
  rate.p = TRUE,
  c = 1.96,
  max.L.m = NA,
  ONUM,
  AGEUNIT,
  cex = 1,
  two.plot = FALSE,
  xlab = "Age (in years)"
)

Arguments

res.nonpar

An object returned by nonpara_esti containing nonparametric estimates.

rate.p

Logical; if TRUE (default), plots marginal rate estimates. If FALSE, plots marginal mean estimates.

c

Numeric; critical value used for constructing pointwise confidence intervals. Default is 1.96 (corresponding to a 95% confidence interval).

max.L.m

Numeric; upper limit for the y-axis. If NA (default), the maximum value is automatically calculated based on the data.

ONUM

Integer; maximum number of time units (i.e., AGEUNIT* number of years).

AGEUNIT

Integer; number of time units per year (e.g., 6 for bimonthly data).

cex

Numeric; magnification factor for legend text size. Default is 1.

two.plot

Logical; if TRUE, splits display into two separate panels, useful when visualizing a large number of covariate combinations. Default is FALSE.

xlab

Character string specifying the x-axis label. Defaults to "Age (in years)".

Value

No return value; generates a plot of marginal mean or rate functions.

References

Chen, A. A., Hu, X. J., & Rosychuk, R. J. (2026). Statistical Learning of Pediatric Mental Health-Related Emergency Department Visits Across COVID-19 Pandemic Periods. arXiv preprint arXiv:2607.26210.

Examples


# The object 'nonpara.res.uk' is the output from nonpara_esti().
nonpara.res.uk <- nonpara_esti(base = my.base[,c(-1,-9)], 
visit = my.visit, mycovars = c("MaleInd","CityAInd","CityBInd"),
  AGEUNIT = 6, ONUM = (18*6), WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), agge.info = my.agge.info,
  seeds = 1234)

# Plot estimated marginal means across covariate subgroups.
plot_nonpar(res.nonpar = nonpara.res.uk, rate.p = FALSE, c = 1.96, ONUM = 108, 
  AGEUNIT = 6, two.plot = FALSE)
  
# Plot estimated marginal rates across covariate subgroups.
plot_nonpar(res.nonpar = nonpara.res.uk, rate.p = TRUE, c = 1.96, ONUM = 108, 
  AGEUNIT = 6, two.plot = FALSE,max.L.m=0.20)

Plot (Cumulative) Baseline Functions from Non-Stratified Cox-Type Models

Description

Displays estimated (cumulative) baseline functions along with pointwise confidence intervals for non-stratified models with both time-varying and time-constant regression coefficients.

Usage

plot_nonstrati_baseline(
  fit.non.s,
  ONUM,
  AGEUNIT,
  subtitle = "non-stratified",
  leg.loc = "topright",
  c = 1.96,
  leg.cex = 1.2,
  cumulative = TRUE,
  ylim = NA,
  cex.lab = 1.3,
  cex.axis = 1.3,
  model.v = "Age-varying coef",
  model.c = "Age-constant coef",
  xlab = "Age (in years)"
)

Arguments

fit.non.s

Object containing the output from a non-stratified model fit (e.g., from nonstrati.reg).

ONUM

Integer; maximum number of time units (i.e., AGEUNIT* number of years).

AGEUNIT

Integer; number of time units per year (e.g., 6 for bimonthly data).

subtitle

Character string specifying plot subtitle. Default is "non-stratified".

leg.loc

Character string indicating legend placement (e.g., "topright").

c

Numeric; critical value used for constructing pointwise confidence intervals. Default is 1.96 (corresponding to a 95% confidence interval).

leg.cex

Numeric magnification factor for legend text size. Default is 1.2.

cumulative

Logical; if TRUE (default), plots estimated cumulative baseline functions; if FALSE, plots estimated baseline functions.

ylim

Numeric vector of length 2 specifying y-axis limits c(ymin, ymax). If NA (default), limits are computed automatically.

cex.lab

Numeric magnification factor for axis labels relative to current setting. Default is 1.3.

cex.axis

Numeric magnification factor for axis tick labels relative to current setting. Default is 1.3.

model.v

Character string; legend label for the model with time-varying coefficients. Default is "Age-varying coef".

model.c

Character string; legend label for the model with time-constant coefficients. Default is "Age-constant coef".

xlab

Character string specifying the x-axis label. Defaults to "Age (in years)".

Details

The function assumes that method = "all" and B!=NA was used in nonstrati.reg.

Value

No return value; generates a (cumulative) baseline function plot.

Examples


# The 'nonstrati.fit.k.B' is the output from nonstrati.reg()

# Note: 95% pointwise confidence intervals for the baseline functions can
# be plotted only when bootstrap- or multiplier-resampling-based variance
# estimates are available from nonstrati.reg().
nonstrati.fit.k.B <- nonstrati.reg(base = my.base[,c(-1,-7,-8)], 
visit = my.visit, 
AGEUNIT = 6, ONUM = 108, 
  WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), 
  etapara = 1e-4, reptmax = 50, windowKH = rep(9,108), 
  tauL = 9, tauR = 105, method = "all", N_star = my.agge.info, n=50000, 
  knownBirth = TRUE, reptw = 100, B = 200, seeds = 12345, resampling = FALSE, 
  r.distn = "std Normal",mycovars = c("MaleInd","CityAInd","CityBInd"), 
  save.res = FALSE, output.lambda = TRUE)

# Plot the estimated cumulative baseline functions:
plot_nonstrati_baseline(fit.non.s = nonstrati.fit.k.B, ONUM = 108, AGEUNIT = 6, 
  subtitle = "non-stratified analysis", cumulative = TRUE, c = 1.96,
  leg.loc="topleft")
  
# Plot the estimated baseline functions:
plot_nonstrati_baseline(fit.non.s = nonstrati.fit.k.B, ONUM = 108, AGEUNIT = 6, 
  subtitle = "non-stratified analysis", cumulative = FALSE, c = 1.96,
  leg.loc="topleft",ylim=c(0.05,0.17))

Plot Estimated Regression Coefficients from Non-Stratified Cox-Type Models

Description

Displays estimates and pointwise confidence intervals for time-varying and time-constant regression coefficients from fitted non-stratified models.

Usage

plot_nonstrati_coef(
  fit.non.s,
  cov_name,
  i = 1,
  age,
  tauL,
  tauR,
  yRange = NA,
  resampling.se = TRUE,
  model.v = "Age-varying coef",
  model.c = "Age-constant coef",
  location = "topright",
  cex.axis = 1.2,
  cex.lab = 1.3,
  cex.legend = 1.2,
  c = 1.96,
  xlab = "Age (in years)"
)

Arguments

fit.non.s

Object containing the output from a non-stratified model fit (e.g., from nonstrati.reg).

cov_name

Character vector of covariate names matching the model specification.

i

Integer; index of the covariate in cov_name to be plotted. Default is 1.

age

Numeric vector representing the age or time points corresponding to the estimates.

tauL

Numeric (in time units); left boundary used for local constant estimation to avoid boundary issues.

tauR

Numeric (in time units); right boundary used for local constant estimation to avoid boundary issues.

yRange

Numeric vector of length 2 specifying the y-axis limits c(ymin, ymax). If NA (default), limits are calculated automatically.

resampling.se

Logical; if TRUE (default), uses standard errors derived from bootstrap/resampling. If FALSE, uses sandwich standard errors.

model.v

Character string; legend label for the model with time-varying coefficients. Default is "Age-varying coef".

model.c

Character string; legend label for the model with time-constant coefficients. Default is "Age-constant coef".

location

Character string specifying the position of the legend (e.g., "topright").

cex.axis

Numeric magnification factor for axis annotation relative to current setting. Default is 1.2.

cex.lab

Numeric magnification factor for axis labels relative to current setting. Default is 1.3.

cex.legend

Numeric magnification factor for legend text. Default is 1.2.

c

Numeric; critical value used for constructing pointwise confidence intervals. Default is 1.96 (corresponding to a 95% confidence interval).

xlab

Character string specifying the x-axis label. Defaults to "Age (in years)".

Details

The function assumes that method = "all" was used in nonstrati.reg.

Value

No return value; generates a coefficient plot for the specified covariate.

Examples


# The 'nonstrati.fit.uk' and 'nonstrati.fit.k' are outputs from nonstrati.reg()
nonstrati.fit.uk <- nonstrati.reg(base = my.base[,c(-1,-9)], visit = my.visit, 
AGEUNIT = 6, ONUM = 108, 
  WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), 
  etapara = 1e-4, reptmax = 50, windowKH = rep(9,108), 
  tauL = 9, tauR = 105, method = "all", N_star = my.agge.info, n=50000, 
  knownBirth = FALSE, reptw = 100, B = NA, seeds = 12345, resampling = TRUE, 
  r.distn = "std Normal",mycovars = c("MaleInd","CityAInd","CityBInd"), 
  save.res = FALSE, output.lambda = TRUE)
  
# 95% pointwise confidence intervals using sandwich-form standard errors.
plot_nonstrati_coef(fit.non.s = nonstrati.fit.uk, cov_name = c("MaleInd","CityAInd","CityBInd"), 
i = 1, age = c(1:108)/6, tauL = 9, tauR = 105,resampling.se = FALSE, c = 1.96)

plot_nonstrati_coef(fit.non.s = nonstrati.fit.uk, cov_name = c("MaleInd","CityAInd","CityBInd"), 
i = 2, age = c(1:108)/6, tauL = 9, tauR = 105,resampling.se = FALSE, c = 1.96)

plot_nonstrati_coef(fit.non.s = nonstrati.fit.uk, cov_name = c("MaleInd","CityAInd","CityBInd"), 
i = 3, age = c(1:108)/6, tauL = 9, tauR = 105,resampling.se = FALSE, c = 1.96)

# 95% pointwise confidence intervals using bootstrap- or
# multiplier-resampling-based standard errors.
nonstrati.fit.k <- nonstrati.reg(base = my.base[,c(-1,-7,-8)], 
visit = my.visit, 
AGEUNIT = 6, ONUM = 108, 
  WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), 
  etapara = 1e-4, reptmax = 50, windowKH = rep(9,108), 
  tauL = 9, tauR = 105, method = "all", N_star = my.agge.info, n=50000, 
  knownBirth = TRUE, reptw = 100, B = 100, seeds = 12345, resampling = TRUE, 
  r.distn = "std Normal",mycovars = c("MaleInd","CityAInd","CityBInd"), 
  save.res = FALSE, output.lambda = TRUE)
  
plot_nonstrati_coef(fit.non.s = nonstrati.fit.k, cov_name = c("MaleInd","CityAInd","CityBInd"), 
i = 1, age = c(1:108)/6, tauL = 9, tauR = 105,resampling.se = TRUE, c = 1.96)

Plot (Cumulative) Baseline Functions from Stratified Cox-Type Models

Description

Displays estimated (cumulative) baseline functions along with pointwise confidence intervals for stratified models with both time-varying and time-constant regression coefficients.

Usage

plot_stratified_baseline(
  fit,
  ONUM,
  AGEUNIT,
  subtitle = "stratified",
  leg.loc = "topright",
  c = 1.96,
  leg.cex = 1.2,
  cumulative = TRUE,
  ylim = NA,
  comLamb = FALSE,
  cex.lab = 1.3,
  cex.axis = 1.3,
  model.v = "Age-varying coef",
  model.c = "Age-constant coef",
  xlab = "Age (in years)"
)

Arguments

fit

Object containing output from a stratified Cox-type model fit (e.g., from strati.reg).

ONUM

Integer; maximum number of time units (i.e., AGEUNIT* number of years).

AGEUNIT

Integer; number of time units per year (e.g., 6 for bimonthly data).

subtitle

Character string specifying plot subtitle. Default is "stratified".

leg.loc

Character string indicating legend placement (e.g., "topright").

c

Numeric; critical value used for constructing pointwise confidence intervals. Default is 1.96 (corresponding to a 95% confidence interval).

leg.cex

Numeric magnification factor for legend text size. Default is 1.2.

cumulative

Logical; if TRUE (default), plots estimated cumulative baseline functions; if FALSE, plots estimated baseline functions.

ylim

Numeric vector of length 2 specifying y-axis limits c(ymin, ymax). If NA (default), limits are computed automatically.

comLamb

Logical; indicates whether common baseline functions across strata were assumed in the model fit. Default is FALSE.

cex.lab

Numeric magnification factor for axis labels relative to current setting. Default is 1.3.

cex.axis

Numeric magnification factor for axis tick labels relative to current setting. Default is 1.3.

model.v

Character string; legend label for the model with time-varying coefficients. Default is "Age-varying coef".

model.c

Character string; legend label for the model with time-constant coefficients. Default is "Age-constant coef".

xlab

Character string specifying the x-axis label. Defaults to "Age (in years)".

Details

The function assumes that method = "all" and B!=NA was used in strati.reg.

Value

No return value; generates a plot of stratum-specific baseline functions.

References

Chen, A. A., Hu, X. J., & Rosychuk, R. J. (2026a). Stratified regression analysis of zero-truncated recurrent event data. The Annals of Applied Statistics, 20(1), 434-451.

Chen, A. A., Hu, X. J., Rosychuk, R. J., & Zeng, L. (2026b). Analyzing zero‐truncated recurrent event data by stratified regression with time‐varying coefficients. Canadian Journal of Statistics, e70064.

Examples


# The 'strati.fit.k.R.p' is the output from strati.reg().
strati.fit.k.R.p <- strati.reg(base = my.base[,c(-1,-7,-8)], visit = my.visit, 
AGEUNIT = 6,ONUM = 108, WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), etapara = 1e-4, reptmax = 50,
  windowKH = rep(9,108), tauL = 9, tauR = 105, method = "all",
  N_star =  my.agge.info, n=50000,
  B = 200, seeds = 12345, com.coef = 0, com.lamb = 0, knownBirth = TRUE, 
  reptw = 100, resampling = TRUE,r.distn = "Poisson",
  mycovars = c("MaleInd","CityAInd","CityBInd"), save.res = FALSE, 
  output.lambda = TRUE)
  
# Plot stratified cumulative baseline function estimates:
  plot_stratified_baseline(fit = strati.fit.k.R.p, ONUM = 108, AGEUNIT = 6, 
  subtitle = "stratified analysis", leg.loc="topleft",cumulative = TRUE, 
  comLamb = FALSE,c = 1.96)
  
  # Plot stratified baseline function estimates:
  plot_stratified_baseline(fit = strati.fit.k.R.p, ONUM = 108, AGEUNIT = 6, 
  subtitle = "stratified analysis", leg.loc="topleft",cumulative = FALSE, 
  comLamb = FALSE,c = 1.96)
  
  # The object 'strati.fit.k.R.p.combB' is the output from strati.reg().
  strati.fit.k.R.p.combB <- strati.reg(base = my.base[,c(-1,-7,-8)], visit = my.visit, 
AGEUNIT = 6,ONUM = 108, WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), etapara = 1e-4, reptmax = 50,
  windowKH = rep(9,108), tauL = 9, tauR = 105, method = "all",
  N_star =  my.agge.info, n=50000,
  B = 200, seeds = 12345, com.coef = 0, com.lamb = 1, knownBirth = TRUE, 
  reptw = 100, resampling = TRUE,r.distn = "Poisson",
  mycovars = c("MaleInd","CityAInd","CityBInd"), save.res = FALSE, 
  output.lambda = TRUE)
  
  # Plot the common cumulative baseline function estimates.
  plot_stratified_baseline(fit = strati.fit.k.R.p.combB, ONUM = 108, AGEUNIT = 6, 
  subtitle = "stratified analysis", leg.loc="topleft",cumulative = TRUE, 
  comLamb = TRUE,c = 1.96)
  
  
  # Plot the common baseline function estimates.
  plot_stratified_baseline(fit = strati.fit.k.R.p.combB, ONUM = 108, AGEUNIT = 6, 
  subtitle = "stratified analysis", leg.loc="topleft",cumulative = FALSE, 
  comLamb = TRUE,c = 1.96)

Plot Estimated Regression Coefficients from Stratified Cox-Type Models

Description

Displays estimates and pointwise confidence intervals for time-varying and time-constant regression coefficients from fitted stratified models.

Usage

plot_stratified_coef(
  fit,
  cov_name,
  i = 1,
  age,
  tauL,
  tauR,
  yRange = NA,
  adjust.LC = FALSE,
  tauR.1 = NA,
  tauL.2 = NA,
  comCoef = FALSE,
  model.v = "Age-varying coef",
  model.c = "Age-constant coef",
  location = "topright",
  cex.axis = 1.2,
  cex.lab = 1.3,
  cex.legend = 1.2,
  c = 1.96,
  xlab = "Age (in years)"
)

Arguments

fit

Object containing output from a stratified Cox-type model fit (e.g., from strati.reg).

cov_name

Character vector of covariate names matching the model specification.

i

Integer; index of the covariate in cov_name to be plotted. Default is 1.

age

Numeric vector representing the age or time points corresponding to the estimates.

tauL

Numeric (in time units); left boundary used for local constant estimation to avoid boundary issues.

tauR

Numeric (in time units); right boundary used for local constant estimation to avoid boundary issues.

yRange

Numeric vector of length 2 specifying the y-axis limits c(ymin, ymax). If NA (default), limits are calculated automatically.

adjust.LC

Logical; if TRUE, applies boundary adjustment using tauR.1 and tauL.2 for stratum-specific data sparsity. Default is FALSE.

tauR.1

Numeric (in time units); right boundary parameter for stratum 1 when adjust.LC = TRUE.

tauL.2

Numeric (in time units); left boundary parameter for stratum 2 when adjust.LC = TRUE.

comCoef

Logical; indicates whether common regression coefficients across strata were assumed in the model. Default is FALSE.

model.v

Character string; legend label for the model with time-varying coefficients. Default is "Age-varying coef".

model.c

Character string; legend label for the model with time-constant coefficients. Default is "Age-constant coef".

location

Character string specifying the position of the legend (e.g., "topright").

cex.axis

Numeric magnification factor for axis annotation relative to current setting. Default is 1.2.

cex.lab

Numeric magnification factor for axis labels relative to current setting. Default is 1.3.

cex.legend

Numeric magnification factor for legend text. Default is 1.2.

c

Numeric; critical value used for constructing pointwise confidence intervals. Default is 1.96 (corresponding to a 95% confidence interval).

xlab

Character string specifying the x-axis label. Defaults to "Age (in years)".

Details

The function assumes that method = "all" was used in strati.reg.

Value

No return value; generates a stratum-specific coefficient plot for the specified covariate.

References

Chen, A. A., Hu, X. J., & Rosychuk, R. J. (2026a). Stratified regression analysis of zero-truncated recurrent event data. The Annals of Applied Statistics, 20(1), 434-451.

Chen, A. A., Hu, X. J., Rosychuk, R. J., & Zeng, L. (2026b). Analyzing zero‐truncated recurrent event data by stratified regression with time‐varying coefficients. Canadian Journal of Statistics, e70064.

Examples


# The 'strati.fit.k.R.p' is the output from strati.reg().
strati.fit.k.R.p <- strati.reg(base = my.base[,c(-1,-7,-8)], visit = my.visit, 
AGEUNIT = 6,ONUM = 108, WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), etapara = 1e-4, reptmax = 50,
  windowKH = rep(9,108), tauL = 9, tauR = 105, method = "all",
  N_star =  my.agge.info, n=50000,
  B = 200, seeds = 12345, com.coef = 0, com.lamb = 0, knownBirth = TRUE, 
  reptw = 100, resampling = TRUE,r.distn = "Poisson",
  mycovars = c("MaleInd","CityAInd","CityBInd"), save.res = FALSE, 
  output.lambda = TRUE)
  
# Plot age-varying regression coefficient estimates for each stratum.
# The age range is adjusted for the local constant estimates in both strata.
plot_stratified_coef(fit = strati.fit.k.R.p, 
cov_name = c("MaleInd","CityAInd","CityBInd"), 
  i = 1, age = c(1:108)/6, tauL = 9, tauR = 105, yRange=c(-3.5,1),
  adjust.LC = TRUE, tauR.1=105,tauL.2=18,comCoef = FALSE)
  
plot_stratified_coef(fit = strati.fit.k.R.p, 
cov_name = c("MaleInd","CityAInd","CityBInd"), 
i = 2, age = c(1:108)/6, tauL = 9, tauR = 105, yRange=c(-2,1),
adjust.LC = TRUE, tauR.1=105,tauL.2=18,comCoef = FALSE, c = 1.96)

plot_stratified_coef(fit = strati.fit.k.R.p, 
cov_name = c("MaleInd","CityAInd","CityBInd"), 
i = 3, age = c(1:108)/6, tauL = 9, tauR = 105, yRange=c(-2,1),
adjust.LC = TRUE, tauR.1=105,tauL.2=18,comCoef = FALSE,c = 1.96)

# Plot common regression coefficient estimates across strata.
strati.fit.k.R.p.combreg <- strati.reg(base = my.base[,c(-1,-7,-8)], visit = my.visit, 
AGEUNIT = 6,ONUM = 108, WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), etapara = 1e-4, reptmax = 50,
  windowKH = rep(9,108), tauL = 9, tauR = 105, method = "all",
  N_star =  my.agge.info, n=50000,
  B = 200, seeds = 12345, com.coef = 1, com.lamb = 0, knownBirth = TRUE, 
  reptw = 100, resampling = TRUE,r.distn = "Poisson",
  mycovars = c("MaleInd","CityAInd","CityBInd"), save.res = FALSE, 
  output.lambda = TRUE)
  
plot_stratified_coef(fit = strati.fit.k.R.p.combreg, 
cov_name = c("MaleInd","CityAInd","CityBInd"), 
i = 1, age = c(1:108)/6, tauL = 9, tauR = 105, yRange=c(-2.5,-1),
adjust.LC = TRUE, tauR.1=105,tauL.2=18,comCoef = TRUE,c = 1.96)

plot_stratified_coef(fit = strati.fit.k.R.p.combreg, 
cov_name = c("MaleInd","CityAInd","CityBInd"), 
i = 2, age = c(1:108)/6, tauL = 9, tauR = 105, yRange=c(-1.4,-0.5),
adjust.LC = TRUE, tauR.1=105,tauL.2=18,comCoef = TRUE,c = 1.96)

plot_stratified_coef(fit = strati.fit.k.R.p.combreg, 
cov_name = c("MaleInd","CityAInd","CityBInd"), 
i = 3, age = c(1:108)/6, tauL = 9, tauR = 105, yRange=c(-2,-1),
adjust.LC = TRUE, tauR.1=105,tauL.2=18,comCoef = TRUE,c = 1.96)

Stratified Regression Analysis for Zero-Truncated Recurrent Event Data

Description

Fits a stratified Cox-type model with time-constant or time-varying regression coefficients using zero-truncated recurrent event data integrated with population census information. The method accounts for coarsened follow-up information and providing bootstrap or multiplier resampling variance estimates.

Usage

strati.reg(
  base,
  visit,
  AGEUNIT,
  ONUM,
  WINDOWL,
  WINDOWR,
  etapara,
  reptmax,
  windowKH,
  tauL,
  tauR,
  method,
  adjust.LC = FALSE,
  tauR.1,
  tauL.2,
  N_star,
  n,
  B = NA,
  seeds,
  com.coef = 0,
  com.lamb = 0,
  knownBirth = FALSE,
  reptw = 100,
  resampling = FALSE,
  r.distn = "std Normal",
  mycovars,
  save.res = FALSE,
  resultpath = NULL,
  output.lambda = FALSE
)

Arguments

base

Matrix of subject-level information. Each row corresponds to one subject and contains demographic information (covariates and either the exact birthdate (birthJulian) or birthdate interval (birthl and birthr)), the number of observed events (nmvst), and the row index of the subject's first observed event in the visit matrix (p1stvt). The variables must be provided in the following order: covariates, nmvst, p1stvt, and either birthJulian or birthl and birthr.

visit

Matrix of event-level information. Each row corresponds to one observed event and contains the subject ID (patid), an indicator of whether the event is a subsequent visit (reptind; 0 for the first observed visit and 1 for subsequent visits for the same subject), the Julian date of the event (date), and the integer age at the time of the event (age). Subject IDs must be integers from 1 through the number of rows in the base matrix. If birthdate information is unavailable, the integer age (age) must be provided for each event.

AGEUNIT

Integer; the number of time units per year (e.g., 6 for bimonthly data).

ONUM

Integer; the maximum number of time units under observation (e.g., 18 * AGEUNIT for <18 years).

WINDOWL

Numeric; Julian date representing the left boundary of the data extraction window.

WINDOWR

Numeric; Julian date representing the right boundary of the data extraction window.

etapara

Numeric; tolerance parameter controlling the convergence criterion when solving the estimating equations.

reptmax

Integer; the maximum number of iterations allowed when solving estimating equations.

windowKH

Numeric; bandwidth size of the smoothing kernel, specified in time units.

tauL

Numeric (in time units); left boundary used for local constant estimation to avoid boundary issues.

tauR

Numeric (in time units); right boundary used for local constant estimation to avoid boundary issues.

method

Character string specifying the estimation method. Options are estimation under the Andersen-Gill model ("L"), local constant estimation under a Cox-type model with time-varying coefficients ("LC"), or both estimation procedures ("all").

adjust.LC

Logical; if TRUE, applies boundary adjustment using tauR.1 and tauL.2 for stratum-specific data sparsity. Default is FALSE.

tauR.1

Numeric (in time units); right boundary parameter for stratum 1 when adjust.LC = TRUE.

tauL.2

Numeric (in time units); left boundary parameter for stratum 2 when adjust.LC = TRUE.

N_star

Data frame containing aggregated population census data, including integer age (AGE),mycovars, and number of subjects in that subgroup (pop).

n

Integer; the approximate size of $O$, the random sample drawn from the target population.

B

Integer or NA; number of bootstrap or multiplier-resampling replicates used for variance estimation. Defaults to NA, in which case no variance estimator is calculated.

seeds

Numeric; seed value for reproducible sampling. Default is 1234.

com.coef

Binary; 0 (default) allows stratum-specific regression coefficients, while 1 constrains coefficients to be common across strata.

com.lamb

Binary; 0 (default) allows stratum-specific baseline functions, while 1 constrains baselines to be common across strata.

knownBirth

Logical; indicates whether exact birthdates are known. Defaults to FALSE. If FALSE, exact birthdates are unknown, and birth.interval can be used to calculate the corresponding birthdate intervals in Julian dates.

reptw

Integer; number of birthdate sets generated for the missing birthdate procedure A (Hu & Rosychuk, 2016). Default is 100.

resampling

Logical; if TRUE, applies the multiplier resampling method described in Chen, Hu, & Rosychuk (2026). If FALSE (default), standard bootstrapping is used.

r.distn

Character string specifying the multiplier distribution used for resampling when resampling = TRUE. Options are the standard normal distribution ("std Normal"; default) or a Poisson distribution with variance 1 ("Poisson").

mycovars

Character vector specifying the variable names of covariates to define subgroups.

save.res

Logical; if TRUE, writes output files to disk. Default is FALSE.

resultpath

Character string indicating the file path destination for saved results, such as "C:\project\". Required when save.res = TRUE.

output.lambda

Logical; if TRUE, includes baseline function estimates in the returned output. Default is FALSE.

Details

The detailed estimation procedures are described in Chen, Hu & Rosychuk (2026a) for time-constant regression coefficients and in Chen et al. (2026b) for time-varying regression coefficients. The approach accounts for missing birthdates using Procedure A of Hu & Rosychuk (2016).

Value

A list containing estimated stratified regression coefficients, standard error estimates (bootstrap- or resampling-based), estimated stratified cumulative baseline functions, and estimated stratified baseline functions when output.lambda = TRUE.

References

Chen, A. A., Hu, X. J., & Rosychuk, R. J. (2026a). Stratified regression analysis of zero-truncated recurrent event data. The Annals of Applied Statistics, 20(1), 434-451.

Chen, A. A., Hu, X. J., Rosychuk, R. J., & Zeng, L. (2026b). Analyzing zero‐truncated recurrent event data by stratified regression with time‐varying coefficients. Canadian Journal of Statistics, e70064.

Hu, X. J., & Rosychuk, R. J. (2016). Imputation procedures for missing birthdates in administrative health data. Statistics in Medicine, 35(14), 2341-2355.

Examples


# When birthdates are unknown, obtain resampling-based variance estimates
# using standard normal multipliers..
strati.fit.uk.R <- strati.reg(base = my.base[,c(-1,-9)], visit = my.visit, 
AGEUNIT = 6,ONUM = 108, WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), etapara = 1e-4, reptmax = 50,
  windowKH = rep(9,108), tauL = 9, tauR = 105, method = "all",
  N_star =  my.agge.info, n=50000,
  B = 200, seeds = 12345, com.coef = 0, com.lamb = 0, knownBirth = FALSE, 
  reptw = 100, resampling = TRUE,r.distn = "std Normal",
  mycovars = c("MaleInd","CityAInd","CityBInd"), save.res = FALSE, 
  output.lambda = TRUE)
  
  # When birthdates are known, obtain resampling-based variance estimates
  # using Poisson multipliers with variance 1.
  strati.fit.k.R.p <- strati.reg(base = my.base[,c(-1,-7,-8)], visit = my.visit, 
AGEUNIT = 6,ONUM = 108, WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), etapara = 1e-4, reptmax = 50,
  windowKH = rep(9,108), tauL = 9, tauR = 105, method = "all",
  N_star =  my.agge.info, n=50000,
  B = 200, seeds = 12345, com.coef = 0, com.lamb = 0, knownBirth = TRUE, 
  reptw = 100, resampling = TRUE,r.distn = "Poisson",
  mycovars = c("MaleInd","CityAInd","CityBInd"), save.res = FALSE, 
  output.lambda = TRUE)
  
  # When birthdates are known, obtain bootstrap-based variance estimates.
  strati.fit.k.B <- strati.reg(base = my.base[,c(-1,-7,-8)], visit = my.visit, 
AGEUNIT = 6,ONUM = 108, WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), etapara = 1e-4, reptmax = 50,
  windowKH = rep(9,108), tauL = 9, tauR = 105, method = "all",
  N_star =  my.agge.info, n=50000,
  B = 200, seeds = 12345, com.coef = 0, com.lamb = 0, knownBirth = TRUE, 
  reptw = 100, resampling = FALSE,r.distn = "Poisson",
  mycovars = c("MaleInd","CityAInd","CityBInd"), save.res = FALSE, 
  output.lambda = TRUE)
  
  # Fit models with common regression coefficients across strata.
  strati.fit.k.R.p.combreg <- strati.reg(base = my.base[,c(-1,-7,-8)], visit = my.visit, 
AGEUNIT = 6,ONUM = 108, WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), etapara = 1e-4, reptmax = 50,
  windowKH = rep(9,108), tauL = 9, tauR = 105, method = "all",
  N_star =  my.agge.info, n=50000,
  B = 200, seeds = 12345, com.coef = 1, com.lamb = 0, knownBirth = TRUE, 
  reptw = 100, resampling = TRUE,r.distn = "Poisson",
  mycovars = c("MaleInd","CityAInd","CityBInd"), save.res = FALSE, 
  output.lambda = TRUE)
  
  # Fit models with a common baseline function across strata.
  strati.fit.k.R.p.combB <- strati.reg(base = my.base[,c(-1,-7,-8)], visit = my.visit, 
AGEUNIT = 6,ONUM = 108, WINDOWL = as.numeric(as.Date("2010-04-01")), 
  WINDOWR = as.numeric(as.Date("2025-03-31")), etapara = 1e-4, reptmax = 50,
  windowKH = rep(9,108), tauL = 9, tauR = 105, method = "all",
  N_star =  my.agge.info, n=50000,
  B = 200, seeds = 12345, com.coef = 0, com.lamb = 1, knownBirth = TRUE, 
  reptw = 100, resampling = TRUE,r.distn = "Poisson",
  mycovars = c("MaleInd","CityAInd","CityBInd"), save.res = FALSE, 
  output.lambda = TRUE)