NEWS

---

trtswitch 0.2.7

• fixed a typo in tsegest.R
• fixed a typo in the settings output of phregr
• fixed a bug in plot.rpsftm, plot.ipe, plot.tsesimp, plot.tsegest, plot.ipcw, and plot.msm for KM plot number at risk tables
• added a check for an empty input vector in the nscpp call in ipcw, msm, and tsegest
• fixed a bug in bootstrap for tsegest by adding bootstrapped values for psi
• added the missing append of oidc_local in bootstrap for tsesimp
• fixed a typo in bootstrap mean_sd for psi for the experimental group in tsesimp
• included only subjects with positive observable follow-up in tssim
• returned an empty matrix if ncol == 0 in splines.cpp
• checked signed K before converting to unsigned K in splines.cpp
• handled an edge case when all 'x' values lie outsided boundary knots in splines.cpp
• set the anyWarning flag to true if either lrEq0 or lrEq1 holds in splines.cpp
• handled an edge case when a stratum does not contain both treatment groups in kmdiff

trtswitch 0.2.6 (2026-05-26)

• add pkgdown workflow for building and deploying the package website to GitHub Pages
• replaced data1[condition, tstop] <- data1[condition, swtrt_time] with data1[condition, "tstop"] <- data1[condition, swtrt_time] in ipcw.R and msm.R because "tstop"" is the variable name in data_switch
• add nthreads for preptdc
• reset number of threads for RcppParallel upon exit from rpsftm, ipe, tsesimp, tsegest, ipcw, and msm.

trtswitch 0.2.5 (2026-03-13)

• replace std::mt19937 with boost::random::mt19937 for random number generation in the bootstrap procedures

trtswitch 0.2.4 (2026-02-12)

• updated the concat_flatmatrix function for empty flat matrices
• utilized column-major storage of flat matrices for cholesky decomposition
• simplified the bygroup function

trtswitch 0.2.3 (2026-01-10)

• added assess_phregr to assess the proportional hazards assumption for the Cox model using the supremum test of Lin et al. (1993)
• added an internal data pbc for testing assess_phregr
• updated the censoring of post switch survival times in ipcw and msm to be consistent with the swtrt_control_only parameter value
• added zph_phregr to compute the Schoenfeld residuals and test statistics for the proportional hazards assumption for the Cox model
• refactored the code in pure c++ and RcppParallel for core computations to improve computational speed

trtswitch 0.2.2 (2025-11-16)

• added a check for the class of object for residuals_lieferegr.R, residuals_phregr.R, and survfit_phregr.R
• added psi_roots for the vector of \code{psi} values at which the Z-statistic is zero, identified using grid search and linear interpolation for rpsftm.
• added psi_roots and psi_trt_roots for the vector of psi values at which the Z-statistic is zero, identified using grid search and linear interpolation for tsegest
• separated NumericVector loghrs = log(hrhats[ok]) into two statements: NumericVector subset_hrhats = hrhats[ok]; NumericVector loghrs = log(subset_hrhats); to avoid ASAN errors for ipcw.cpp, ipe.cpp, msm.cpp, rpsftm.cpp, tsegest.cpp, and tsesimp.cpp
• updated the lrtest call to add the value for the new weight_readj parameter for ipcw.cpp, ipe.cpp, msm.cpp, rpsftm.cpp, tsegest.cpp, and tsesimp.cpp
• added the weight_readj parameter to lrtest to indicate whether the weight variable at each event time will be readjusted to be proportional to the number at risk by treatment group in survival_analysis.
• updated the getpsiest function to output multiple roots depending on the new direction parameter value in utilities.cpp and utilities.h
• updated rpsftm and tsegest to use gridsearch by default
• added print method for rpsftm, ipe, tsesimp, tsegest, ipcw, and msm
• added class attribute for rpsftm, ipe, tsesimp, tsegest, ipcw, and msm objects
• added km_outcome and lr_outcome for rpsftm, ipe, tsesimp, tsegest, ipcw, and msm objects
• clarified the point estimate of psi when there are multiple roots and grid search is used
• initialized weights to 1.0 for the Cox switching model in ipcw
• added event summary and weight summary to output for ipcw and msm
• added event summary to output for rpsftm, ipe, tsesimp and tsegest
• added the plot method for rpsftm, ipe, tsesimp, tsegest, ipcw, and msm
• use settings to store input parameter values for logisregr, liferegr, phregr, rpsftm, ipe, tsesimp, tsegest, ipcw, and msm
• converted treatment back to a factor variable for output data sets in rpsftm, ipe, tsesimp, tsegest, ipcw, and msm
• add res_aft to ipe and tsesimp

trtswitch 0.2.1 (2025-11-06)

• fixed the bootstrap procedures for IPCW, MSM, and TSEgest by reconstructing idx, idx1, treatn1, and stratumn1
• clarified the distinction between IPCW using pooled logistic regression and IPCW using a Cox model with time-dependent covariates
• excluded existing variables in out$data_switch[[h]]$data from add_vars for IPCW and MSM
• excluded existing variables in out$analysis_switch$data_logis[[h]]$data from add_vars for TSEgest
• excluded existing variables in out$data_aft[[h]]$data from add_vars for TSEsimp
• replaced chained sorting using logical OR with multiple if–return statements to ensure stable sorting in survival_analysis
• replaced chained sorting using logical OR with std::tie in IPCW, IPE, MSM, RPSFTM, TSEsimp, and TSEgest
• added id = "id" in test-ipe.R and test-rpsftm.R
• combined functions f and g into a single function in test-rpsftm.R
• added bootstrap testing for test-tsesimp.R
• removed the target argument from the definition of function f in test-tsegest.R
• added an option to use a Cox model with robust variance for constructing confidence intervals for the hazard ratio in IPCW and MSM
• removed robust variance from the switching model in IPCW and MSM to improve computational speed
• replaced geom_density with geom_histogram in the IPCW and MSM vignettes
• removed natural spline basis functions for time from the pooled logistic regression model equation in TSEgest and incorporated these effects into time-dependent confounders in the TSEgest vignette
• replaced event = "died" with event = "event" in bootstrap testing in the TSEgest vignette
• removed the analysis_switch level from the output of TSEgest and moved its components to the top level of the output list
• updated getpsiest to return only the first root
• updated IPCW to apply weight truncation within each treatment group after computing weights for data_outcome
• updated TSEsimp and TSEgest to use the switch time as the progression time when switching occurs before or in the absence of recorded progression
• updated kmest to add the weight parameter for adjusted Kaplan-Meier estimate by Xie & Liu (2005)
• updated lrtest to add the weight parameter for adjusted log-rank test by Xie & Liu (2005, 2011)
• updated ipcw and msm to add Kaplan-Meier estimates of the survival functions and the log-rank test for the treatment effect based on the weighted outcome data truncated at time of treatment switching

trtswitch 0.2.0 (2025-10-06)

• added MSM to package description
• added the plannedTime argument to the tssim function so that followupTime only refers to the follow-up time in a fixed follow-up design
• added an extra record for a subject to the tssim output dataset if the observed survival time exceeds the number of regular treatment cycles times days per cycle
• removed the getAccrualDurationFromN utility function
• changed the default value of the boot parameter to FALSE for ipcw and msm
• added the ns_df parameter to the tsegest function
• added the preptdc function to prepare survival data with time-dependent covariates
• expanded the findInterval3 function definition to mimic R's findInterval() behavior with rightmost_closed, all_inside, and left_open options
• updated the survfit_phregr function to only include time points between the first tstart and the last tstop for each subject with counting process style of input
• added the exclusion of observations with missing values for survfit_phregr and residuals_phregr
• added the exclusion of observations with missing values for rpsftm, ipe, tsesimp, tsegest, ipcw, and msm
• changed to sort the input data by treatment, ID, and time variables for IPCW, MSM, and TSEgest, and by treatment, stratum, ID, and time variables for bootstrap samples
• changed the default value of treat_alt_interaction from FALSE to TRUE for MSM
• renamed Y to event for tssim output
• added the bisect utility function for root finding
• added the root_finding parameter to rpsftm, ipe, tsegest to allow the user to choose between brent and bisect for root finding

trtswitch 0.1.9 (2025-09-03)

• added fail_boots_data to include the failed bootstrap sample data
• used strata variable names and values for dummy variables in logistic regression and life regression models
• updated the tssim function to allow both fixed and variable follow-up designs
• used stratified bootstrapping for the rpsftm, ipe, tsesimp, tsegest, ipcw, and msm functions
• allowed event, pd, and swtrt to take real values 1 or 0 in rpsftm, ipe, tsesimp, tsegest, ipcw, and msm functions
• added martingale residuals for lifereg for event or right-censored survival data
• added time-dependent covariate visit7on and used cattdc instead of catlag for tsegest sample call
• added interval splitting if treatment switching occurs between visits for ipcw, msm, and tsegest
• added a small number to os_time if needed to ensure pd_time is less than os_time for tsesimp and tsegest
• added getpsiest and getpsiend utility functions for psi estimation of rpsftm, ipe, and tsegest
• updated the condition to include post progression data up to switching in tsegest
• updated the condition for time-dependent crossover indicator in tsegest
• updated the algorithm for generating timePFSobs in tsegestsim
• removed catlag and xotime_upper from the output data set of tsegestsim
• applied weights to the next interval for ipcw and msm using logistic regression switching model
• updated tssim to use Z and A to denote the alternative therapy status at the start instead of end of interval
• updated logisregr, liferegr and phregr to allow zero events to run with trivial output

trtswitch 0.1.8 (2025-07-12)

• eliminated redundant code for the logistic regression switching model for ipcw
• added the marginal structural model (msm) method
• added na.action = na.pass for model frame construction involving covariates for all methods
• added the init parameter and the fail flag to the output to logisregr, liferegr, and phregr
• replaced the survreg initial value method with the OLS initial value method for liferegr
• added the fail flag to output of rpsftm, ipe, tsesimp, tsegest, ipcw, and msm

trtswitch 0.1.7 (2025-06-05)

• updated survival_analysis to ignore intervals not at risk within each stratum without creating non overlapping times across strata
• updated documentation for the survsplit utility function
• updated tsegestsim to use the standard definition of weibull scale
• removed bc from logisregr
• removed swtrt_time_upper from tsegest
• removed subject-level adsl data from tsegestsim output
• removed swtrt_time_lower, swtrt_time_upper, and relative_time from ipcw
• removed the robust option for logistic regression treatment switching model in ipcw
• added match3 utility function to match on id and value
• added residuals_liferegr for residuals from parameteric regression models for failure time data
• added the psi_test, aft_dist, and strata_main_effect_only parameters to rpsftm to allow parametric regression and cox regression models to estimate psi
• added baseline covariates to the output Sstar data set of the rpsftm and ipe functions
• added a check for switching time before progression time in tsesimp and tsegest
• added a check for switching time or progression time less than offset in tsesimp and tsegest
• added timeOS and died variables for subject-level data and used event for death status at the end of each time interval for tsegestsim
• added tssim for treatment switching data simulation

trtswitch 0.1.6 (2025-04-25)

• fixed an ASAN issue in survQuantile

trtswitch 0.1.5 (2025-04-24)

• updated the default search interval for psi to [-2,2] at 101 points in rpsftm
• added parameters and output low_psi = -2 and hi_psi = 2 to ipe
• renamed gest to est_psi_tsegest and updated the default search interval for psi to [-2,2] at 101 points in tsegest
• replaced dummy variables with original variables in data_outcome for rpsftm
• replaced dummy variables with original variables in data_outcome and data_aft for ipe
• removed dummy strata variables in data_aft for ipe
• replaced dummy variables with original variables in data_outcome and data_aft for tsesimp
• removed dummy strata variables in data_aft for tsesimp
• renamed time to pps in data_aft for tsesimp
• added pd_time, swtrt_time, and time to data_aft for tsesimp
• added tstart, tstop to data_logis for tsegest
• replaced dummy variables with original variables in data_outcome and data_logis for tsegest
• removed dummy strata variables in data_logis for tsegest
• added tstart, tstop to data_switch for ipcw
• replaced dummy variables with original variables in data_outcome and data_switch for ipcw
• removed dummy strata variables in data_switch for ipcw
• updated liferegr to use better starting values for model parameters
• updated documentations for rpsftm, ipe, tsesimp, tsegest, and ipcw to clarify the variables in output data frames
• added gridsearch parameter to tsegest
• updated unit tests and vignettes to use functions from the survival package
• added maxiter and eps to logisregr, liferegr, and phregr
• added the special case for psilower and psiupper when no root exists when using uniroot finding in rpsftm and tsegest
• updated to shortening survival if patients switched from active treatment to control in tsegestsim
• added subject-level data to tsegestsim

trtswitch 0.1.4 (2025-03-20)

• Added the keep_censor parameter to the kmest function and added the ncensor variable to the output data frame.
• Added time 0 and the ncensor variable to the output data frame of the kmest function
• Added time 0 and the ncensor variable to the output baseline hazard data frame of the phregr function
• Added id to output data

trtswitch 0.1.3 (2025-02-06)

• Updated logistic regression and associated programs

trtswitch 0.1.2 (2024-12-05)

• Added probit and complementary log-log links to logistic regression
• Removed verbatim environment for ggplot in vignettes

trtswitch 0.1.1 (2024-11-01)

• Fixed clang-UBSAN issues.

• Updated description in trtswitch-package.

• Updates to survival.cpp:

  • Changed requirement of positive time to nonnegative time for kmest, kmtest, rmest, and rmdiff.

• Updates to tsegestsimp.cpp:

  • Changed the description of shape1, scale1, shape2, and scale2 parameters.

• Updates to utilities.cpp:

  • Moved untreated and unswitched functions here. These functions are shared by rpsftm and ipe.

• Updates to rpsftm:

  • Added treat variable to Sstar.
  • Changed the default value of admin_recensor_only from FALSE to TRUE.
  • Added data_outcome and fit_outcome to output.
  • Added raw stratum and treat information to data_outcome.

• Updates to ipe:

  • Added treat variable to Sstar.
  • Changed the default value of admin_recensor_only from FALSE to TRUE.
  • Modified Sstar and kmstar to be consistent with rpsftm.
  • Added data_aft and fit_aft to output.
  • Added data_outcome and fit_outcome to output.
  • Added raw stratum and treat information to data_outcome.

• Updates to tsesimp:

  • Changed the default value of admin_recensor_only from FALSE to TRUE.
  • Added data_aft, fit_aft, data_outcome, and fit_outcome to output.
  • Added psi1hats in bootstrap.
  • Added raw stratum information to data_outcome.
  • Added agerand, sex.f, tt_Lnum, rmh_alea.c, and pathway.f as baseline covariates in the example.

• Updates to tsegest:

  • Changed the default value of admin_recensor_only from FALSE to TRUE.
  • Added the parameter n_eval_z to evaluate the Wald statistics at a sequence of psi values.
  • Added data_switch, km_switch, eval_z, data_nullcox, fit_nullcox, data_logis, fit_logis, data_outcome, and fit_outcome to output.
  • Included data_switch, km_switch, eval_z, data_nullcox, fit_nullcox, data_logis, and fit_logis in the analysis_switch list.
  • Added raw stratum information to data_outcome.

• Updates to ipcw:

  • Added relative_time parameter.
  • Added data_switch to output.
  • Added ns_df to output.
  • Renamed df_outcome to data_outcome.
  • Added raw stratum information to data_outcome.

• Added vignettes:

  • rpsftm.Rmd
  • ipe.Rmd
  • tsesimp.Rmd
  • tsegest.Rmd
  • ipcw.Rmd

trtswitch 0.1.0 (2024-10-17)

• Initial release.