Changelog

Source: NEWS.md

lrstat 0.3.5

• added a small positive value to avoid zero time in survival_analysis.cpp

lrstat 0.3.42026-09-15

• removed the rankp0 argument and output from the seamless design functions and simulators, which now always select the most promising treatment at the end of phase 2
• added bm_rank and pe_rank methods to lrsim_bmTrtSel for rank-based Dunnett adjustment for dose selection in stage 1 in seamless phase 2/3 designs
• added getNFromNeventsAndFollowup and its C++ backend in lrstat.cpp for accrual duration calculations under target-event and follow-up constraints
• corrected the minimum weighted p-value ratio initialization in fadjpduncpp (used by fadjpdun) from 1.0 to positive infinity so that ratios exceeding 1 are properly accounted for when computing intersection p-values
• added the simulate parameter to the lrschoenfeld function, and added a half-count correction for extreme simulated powers
• added the Bonferroni p-value combination method ctbonferroni to lrsim_bmTrtSel
• added the familywise error rate fwer to the output of lrsim_bmTrtSel
• added disjunctive and conjunctive power to each byMethod result from lrsim_bmTrtSel and to the run_scenario output data frame
• added event-driven phase III analyses to lrsim_bmTrtSel through the totalNumberOfEvents target, while retaining studyDurationPhase3 for fixed-duration analyses
• changed long-term time-to-event endpoint generation in lrsim_bmTrtSel from response-status-stratified hazards to a copula-based approach, with the latent biomarker response variable correlated directly with the TTE endpoint
• optimized the CE variants in lrsim_bmTrtSel by precalculating the sample size based nominal time stage boundaries once before data generation
• optimized selection-boundary calculations by making use of no efficacy stopping in stage 1.
• optimized exitprob_seamless under equal treatment effects by evaluating the phase-2 selected-arm probabilities once and using arm symmetry for the remaining arms
• renamed the fPCStagewise input parameters stg1_inthyp_nr and stg2_elemhyp to stg1_inthyp_nr_idx and stg2_elemhyp_idx, respectively, to make their index semantics explicit and align them with the output names and related adaptive-testing interfaces
• standardized lrsim_bmTrtSel testing method identifiers to lowercase, including inputs, returned methods and byMethod names, and sumdataTTE rejection indicator columns
• added sumdataBIN and sumdataTTE summary datasets to lrsim_bmTrtSel, including zBiomarker in sumdataBIN and all individual and pooled stage 1 log-rank Z statistics in sumdataTTE to allow derivation of rejection decisions for all methods, along with optional subject-level rawdataBIN and rawdataTTE datasets controlled by maxNumberOfRawDatasets
• added rmvnorm and related multivariate-normal sampling helpers for drawing correlated random vectors from a covariance matrix via Cholesky-based generation and seeded overloads
• added pordmvnormr to compute lower- and upper-tail probabilities for order statistics of multivariate normal random variables, with analytic shortcuts for nonnegative compound-symmetry covariance structures and inclusion-exclusion over reduced-dimensional orthant probabilities for general covariance matrices
• removed mvtnorm from Suggests
• changed the default value of corrEfficacyToxicity in lrsim_bmTrtSel_cpp() from 0.5 to 0
• corrected a collapse argument typo in print.nbpower
• consolidated alpha- and beta-spending function formatting across print methods through the shared .format_spending() helper, preserving the existing printed output

lrstat 0.3.32026-08-25

• standardized C++ pointer/reference declaration style across src with a project-level .clang-format (80-column, 2-space indent, type *name and type &name alignment) and aligned headers/sources to that convention

• excluded .clang-format from package build artifacts via .Rbuildignore

• corrected conditional Type I error calculations in adaptDesign, adaptDesign_multiarm, adaptDesign_seamless, getCP, getCP_multiarm, and getCP_seamless to use normalized post-adaptation information rates consistently, including when the Muller and Schafer method is selected

• added finthyp to construct the intersection-hypothesis indicator matrix for a given number of elementary hypotheses, matching the row order used by fwgtmat and fDefaultWgtmat, for use when migrating custom weight matrices to the new fadjpbon/fadjpsim/fadjpdun interfaces

• reduced per-iteration buffer allocation in fadjpsim’s Simes p-value adjustment by reusing preallocated vectors instead of allocating new ones for each family block

• avoided recomputing the transform gradient for each probability in survQuantile, computing it once per event time instead

• simplified graphical and adaptive multiple-testing p-value adjustment interfaces by making p-values the first argument and automatically supplying equal intersection weights, a single family, and default within-family correlations when these inputs are omitted

• added public R wrappers with nthreads controls for the multi-arm and seamless design, conditional power, and confidence interval functions

• simplified one-active-arm getBound_multiarm and getBound_seamless calculations by delegating to getBound

• added rank-adjusted and conditional-error-updated TSSSD methods to lrsim_bmTrtSel (TSSSD.k, TSSSD.uk, TSSSD.k.rank, TSSSD.uk.rank, TSSSD.k.ce, TSSSD.uk.ce, TSSSD.k.rank.ce, TSSSD.uk.rank.ce) and improved TSSSD boundary computation efficiency, including cached single-arm boundaries and precomputed CE nominal boundaries

• replaced the lpSolve dependency in getDesignAgreement with a native C++17 simplex solver exposed through Rcpp, preserving the documented design results and adding regression tests for the agreement constraints

• added lrsim_bmTrtSel and print.lrsim_bmTrtSel for simulation and summary output of phase 2/3 seamless design selecting one dose at the end of phase 2 based on the posterior benefit-risk tradeoff of correlated binary efficacy and toxicity endpoints, and comparing the closed testing procedures with inverse normal combination, the conditional error rate method, the two-stage seamless design boundaries, and the unadjusted tests for the confirmatory analysis of the long-term endpoint

• added unit tests for the adaptive two-stage multiple testing functions

• allowed alpha1 = 0 in fCERStageBound when no stage 1 efficacy stopping is planned

• set the stage 2 boundary to infinity in fCERNewBound when the conditional error rate exceeds 1

• used struct output to reduce overhead of returning multiple values for exitprobcpp, exitprob_seamless_cpp and exitprob_multiarm_cpp

• allowed the retesting of an active hypothesis if its weight has increased due to the rejection of another hypothesis in fseqbon

• added getDesign2 for futility boundary under the null hypothesis for group sequential designs

• renamed mams to multiarm for more informative description

• added rdsim_seamless for simulation of phase 2/3 seamless design for risk difference

• updated lrsim_seamless and rdsim_seamless to use rank-based phase-2 selection naming and outputs, including rankp0 in the wrapper inputs, overview output, and print methods

• updated seamless_design.cpp related functions to use rank-based treatment selection with rankp0 throughout seamless design, conditional power, confidence interval, and print outputs

• added rdsim_multiarm for simulation of multi-arm multi-stage design for risk difference

• added lrsim_mcpmod and print.lrsim_mcpmod for simulation and summary output of MCPMod design using the log-rank test

• updated mvnormr to support positive semidefinite covariance matrix via minimal diagonal stabilization during factorization

• used seed = 314159 as the default for pmvnormr and qmvnormr to ensure the same random numbers are generated and results are reproducible across different runs

• added the selection to Stage 2 probability in lrsim_seamless

• fixed liferegr to handle empty optional variable inputs (stratum, covariates, weight, offset, id) without triggering a formula parse error from an empty right-hand side

• updated fadjpbon, fadjpsim, and fadjpdun to add the local p-values and the incidence matrix for the intersection hypotheses

• updated fadjpdun for calculating the local p-values for the intersection hypotheses

• updated ftrunc, fstdmix, and fmodmix to return a list with padj and pinter components instead of a bare matrix/vector of adjusted p-values

lrstat 0.3.22026-05-13

• removed the pivoting step for pmvnormr and qmvnormr to take advantage of independent increment structure of score process
• removed the fast parameter for pmvnormr and qmvnormr so that the default is to use fast approximation
• updated the point estimate for getRCI
• renamed tsssd to seamless in the function names for two-stage seamless sequential design for treatment/dose selection
• updated the critical values when typeAlphaSpending = “none” in getDesign_seamless
• added integratedTrial to adaptDesign
• updated the backward image confidence interval method to account for the case when astar > conditional_power
• updated point estimate for repeated confidence interval
• added functions for multi-arm multi-stage (MAMS) design and analysis, including exitprob_multiarm, getBound_multiarm, getDesign_multiarm, adaptDesign_multiarm, getCI_multiarm, getADCI_multiarm
• ensured that if typeAlphaSpending is “OF”, “P”, “WT”, or “none”, then informationRates, efficacyStopping, and spendingTime must be of full length kMax, and informationRates and spendingTime must end with 1 for getBound, getBound_seamless, and getBound_multiarm
• ensured that if typeAlphaSpending is “OF”, “P”, “WT”, or “none”, then informationRates, efficacyStopping, and spendingTime must be of full length kMax, and informationRates and spendingTime must end with 1 for getCI, getRCI
• ensured that if typeAlphaSpendingNew is “OF”, “P”, “WT”, or “none”, then informationRatesNew, efficacyStoppingNew, and spendingTimeNew must be of full length kNew, and informationRatesNew and spendingTimeNew must end with 1 for getADCI, getADRCI
• moved up the position of parameter MullerSchafer in getADCI, getADRCI
• added getCP_seamless for conditional power calculation for two-stage seamless sequential design for treatment/dose selection
• added getCP_multiarm for conditional power calculation for multi-arm multi-stage design
• added adaptDesign_seamless for power and sample size calculation of adaptive two-stage seamless sequential design for treatment/dose selection
• added getCI_seamless for confidence interval calculation for two-stage seamless sequential design for treatment/dose selection
• added getADCI_seamless for confidence interval calculation using the backward image method for adaptive two-stage seamless sequential design for treatment/dose selection
• moved the f_pvalue function from confidence_interval.cpp to generic_design.cpp
• updated infoRatesNew when informationRatesNew is not missing in adaptDesign_seamless_cpp and adaptDesign_multiarm_cpp
• added default value of 1 for parameter r of adaptDesign_seamless
• updated pmvnormmccpp to use s and T instead of Ivec
• added byLevelBounds in the output of getDesign_multiarm
• updated the condition for calculating cpu0 in adaptDesign_multiarm_cpp
• updated the description for adaptDesign_multiarm output
• updated the default value of r and rNew to 1 in adaptDesign_multiarm and getADCI_multiarm
• updated the stage‑1 covariance used for bound calculation in getBound_seamless to account for corr_known when using alpha‑spending functions
• used memoization to avoid recalculating p0 and p1 in the lambda h of preject_by_arm (exitprob_seamless)
• added stopStage to output data sets and added bestArm and reject to sumdata2 output data set for lrsim_seamless
• used NULL as the default value for lambdas and gammas for lrsim_seamless
• added lrsim_multiarm for simulation of multi-arm multi-stage design
• added futilityBounds, futilityCP, futilityTheta, futilityBoundsInt, futilityCPInt, futilityThetaInt to adaptDesign, adaptDesign_seamless, and adaptDesign_multiarm
• removed typeBetaSpending and parameterBetaSpending from adaptDesign, adaptDesign_seamless, and adaptDesign_multiarm

lrstat 0.3.12026-04-08

• calculated eventsPerStage, dropoutPerStage, subjectsPerStage, and analysisTimePerstage in the lrsim output based on trials having the stage (i.e., did not stop before the stage).
• added mvnormr for multivariate normal probability and quantile calculations
• added tsssd_design: analytical calculation for two-stage seamless sequential designs for treatment/dose selection.
• added lrstim_tsssd: simulation function for two-stage seamless sequential designs for treatment/dose selection.
• printed power with 4 instead of 3 decimal places in the output
• used std::size_t instead of size_t in all header files
• added pnorm_fast, qnorm_acklam, expand_stratified_to_slice, and invchol utility functions

lrstat 0.3.02026-03-07

• replaced accrualDuration with n for lrsim, lrsim2e, lrsim3a, lrsim2e3a, lrsimsub, and binary_tte_sim to allow the simulation to run with a fixed number of subjects
• added time2 and weight parameters to kmest
• added time2, weight, and weight_readj parameters to lrtest
• added assess_phregr and zph_phregr functions
• updated findInterval3 to add rightmost_closed, all_inside, and left_open parameters
• added dtpwexp for density function of truncated piecewise exponential distribution
• updated hazard_pd to match probabilities at quantiles in addition to cut points
• added corr_pfs_os to obtain correlation between PFS and OS given correlation between PD and OS
• added hazard_pd2 to obtain the hazards for PD to match distribution for PFS and correlation between PFS and OS
• renamed rho to rho_pd_os for added clarity for lrsim2e and lrsim2e3a
• updated hazard_sub to match probabilities at quantiles in addition to cut points
• moved logparams to logistic_regression.cpp and removed logparams, f_der_0, f_ressco_0, logisregloop, and logisregplloop from logistic_regression.h
• updated lrsim2e and lrsim2e3a to account for additional quantile-based cut points for pd
• updated lrsimsub to account for additional quantile-based cut points for biomarker negative sub population
• moved aftparams and coxparams to survival_analysis.cpp and removed aftparams, f_der_1, f_ressco_1, f_jj_1, liferegloop, liferegplloop, f_ld_1, coxparams, f_der_2, phregloop, phregplloop, f_basehaz, f_ressco_2, f_resmart, f_ressch, and f_der_i_2 from survival_analysis.h
• added ptpwexpcpp1 and updated ptpwexpcpp accordingly
• added mtpwexp for the mean and variance of a truncated piecewise exponential distribution
• added quad2d for numerical integration in two dimensions
• removed hd, pd, ad, house, row_house, col_house, givens, row_rot, col_rot, house_bidiag, zero_diagonal, and svd_step from utilities.h
• renamed the parameters in lrsim2e and lrsim2e3a, replacing e1 by pfs and e2 by os
• ran the simulations in serial mode to pass CRAN checks

lrstat 0.2.152025-07-14

• added na.action = na.pass for model frame construction involving covariates for all methods and updated lrstat-package.R to add na.omit and na.pass to reflect this change
• added pbvnorm for the distribution function for standard bivariate normal
• added hazard_pd for the hazard function for progressive disease given the piecewise hazard functions for PFS and OS and the correlation between PD and OS
• added hazard_sub for the hazard function for the biomarker negative subpopulation given the hazard function of the ITT population, the hazard function and the prevalence of the biomarker positive subpopulation
• added the init parameter and the fail flag to the output of logisregr, liferegr, and phregr
• added the nullVariance parameter to getDesignOneProportion
• added details section to the documentation of getDesignPairedPropMcNemar, getDesignRiskDiff, getDesignRiskRatio, getDesignRiskRatioFM, getDesignOddsRatio, getDesignOneMultinom, getDesignTwoMultinom, getDesignTwoOrdinal, getDesignOrderedBinom, getDesignUnorderedBinom, getDesignUnorderedMultinom, pwexploglik, pwexpcuts, and lrschoenfeld
• added references to the documentation of ClopperPearsonCI, BOINTable, mTPI2Table
• added the tol parameter to pwexpcuts
• added the lrsimsub function for log-rank test simulation for enrichment design
• updated lrsim2e and lrsim2e3a to use the hazard_pd function when generating PFS time
• updated kmstat.cpp and rmstat.cpp to report an error message if the information at the first interim is less than the information at the milestone time
• updated lrstat.cpp to avoid the redundant calculation of the final analysis time in the lrpower, lrsamplesize, lrpowerequiv, and lrsamplesizeequiv functions
• updated getDesignMeanDiffCarryover and getDesignMeanDiffCarryoverEquiv to fix the calculation of p and v
• updated getDesignMeanDiff, getDesignMeanDiffXO, and getDesignSlopeDiff to add the parameter value at the null hypothesis to the boundaries for the parameter of interest for noninferiority trials
• updated getDesignPairedMeanDiffEquiv, getDesignMeanDiffEquiv, getDesignMeanDiffXOEquiv, and getDesignSlopeDiffMMRM to simplify the calculation of attainedAlpha for equivalence trials
• updated getDesignWilcoxon, getDesignMeanDiffMMRM, getDesignOneSlope, getDesignSlopeDiff to add details to the documentation
• updated getDesignMeanDiffMMRM to add numberOfCompleters to the output and a reference to the documentation
• updated getDesignMeanDiffCarryover and getDesignMeanDiffCarryoverEquiv to add half_width and nu to the output
• updated getDesignTwoWayANOVA to ensure the sample sizes are multiples of 4
• updated getDesignSlopeDiffMMRM to add more detailed description for the parameters w and N
• updated getDesignSlopeDiffMMRM to add the fixedFollowup parameter to allow fixed follow-up designs and modify the function for computing information and add number of completers accordingly
• updated getDesignSlopeDiffMMRM to add more content to the details section of the documentation
• updated the definition of the Z statistic in the details section of the documentation for getDesignEquiv
• updated print.designMeanDiffMMRM to add numberOfCompleters and analysisTime to the output
• updated print.designMeanDiffCarryover and print.designMeanDiffCarryoverEquiv to add half_width to the output
• updated print.designSlopeDiffMMRM to add numberOfCompleters to the output for fixed follow-up designs
• updated the documentation of the output of lrschoenfeld
• replaced the survreg initial value method with the OLS initial value method for liferegr

lrstat 0.2.142025-06-09

• 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
• removed bc from logisregr
• added residuals_liferegr for residuals from parametric regression models for failure time data
• added maxiter and eps to logisregr, liferegr, and phregr
• added initial values for liferegcpp
• moved survQuantile from getDesignSurvival to survival_analysis.cpp
• moved the detailed implementation from residuals_phregr to residuals_phregcpp
• added detail description of the spending functions for errorSpent

lrstat 0.2.132025-03-20

• checked whether the rounded up n is different from the original n0 before updating accrual intensity or accrual duration
• changed accrualIntensity to accrualIntensity1 in lrsamplesizeequiv when obtaining the timing of interim analysis
• updated kmpower1s and rmpower1s to remove drift and inflation factor from the output while adding number of events, number of dropouts, and number of milestone subjects to the output
• simplified the calculation of underlying survival probability for each treatment group in kmstat1
• updated kmpowersamplesize, kmsamplesize1s, kmsamplesizeequiv, rmpowersamplesize, rmsamplesize1s, rmsamplesizeequiv to remove the impossible case when followupTime is missing for fixed follow-up
• updated kmpowersamplesize, kmsamplesize1s, kmsamplesizeequiv, rmpowersamplesize, rmsamplesize1s, rmsamplesizeequiv to add a check for solving accrualIntensity when study duration is shorter than or equal to milestone
• updated kmpowersamplesize, kmsamplesize1s, kmsamplesizeequiv, rmpowersamplesize, rmsamplesize1s, rmsamplesizeequiv to add a case to decrease accrual duration for fixed follow-up under H0
• updated getDesignMeanDiffCarryover to allow specification of the treatment comparison of interest and whether to account for carryover effects in power calculation
• added getDesignMeanDiffCarryoverEquiv to test for equivalence in mean difference for the direct treatment effect of interest in crossover designs
• added the keep_censor parameter to the kmest function to specify whether to retain censoring time points in the output data frame
• added interval to survsplit output
• used the same stopping criteria for SAS PROC LOGISTIC profile likelihood
• added the float_to_fraction utility function
• renamed runShinyApp to runShinyApp_lrstat

lrstat 0.2.122025-01-17

• allowed time to take zero values in kmest, lrtest, rmest, and rmdiff
• added the bc parameter for bias correction for noncanonical parameterization and weighted logistic regression
• added loop to look for appropriate bracketing interval for the brent algorithm in lrstat, kmstat, rmstat, and nbstat
• updated f_info nbstat to calculate the information directly
• updated the mean exposure calculations in nbstat for starting values for reml
• added svdcpp for singular value decomposition of a rectangular matrix
• added number of events, number of dropouts and number of milestone subjects to kmstat and kmpower
• added number of events, number of dropouts and number of milestone subjects to rmstat and rmpower
• removed variance ratios for equivalence trials assuming Wald test statistics with variance of parameter estimator evaluated under the alternative hypothesis
• removed the interval argument from getDurationFromNevents
• added lrschoenfeld for sample size calculation using the Schoenfeld formula under proportional hazards

lrstat 0.2.112024-12-04

• used numerical integration for lrstat for cases including unequal dropouts between treatment groups, hence removed the numSubintervals argument from lrstat1, lrstat, lrpower, lrsamplesize, lrpowerequiv, and lrsamplesizeequiv
• updated shiny app accordingly
• added noncanonical parametrizations for logisregr

lrstat 0.2.102024-10-18

• used newton-raphson instead of the vmmin algorithm for liferegr and phregr
• removed rpsft with all treatment switching functions moved to trtswitch package
• added the survfit_phregr function to obtain the survivor curve estimates
• added the residuals_phregr function to obtain the martingale, deviance, score, and schoenfeld residuals
• added the shilong dataset
• replaced std::ceil(x) with std::ceil(x - 1.0e-12) to handle superficial precision in binary presentation of a double number
• allowed rep and stratum parameters to represent more than one character variables in the source data
• added the logisregr function for logistic regression of a binary endpoint.

lrstat 0.2.92024-06-23

• fixed the runtime error: nan is outside the range of representable values of type ‘int’
• added the binary_tte_sim function to simulate two endpoints with one being a binary endpoint and the other being a time-to-event endpoint
• added the rpsft function to estimate hazard ratio using rank-preserving structured failure time model

lrstat 0.2.82024-06-19

• updated std::sort handling for Rcpp character vectors
• used static cast to explicitly cast “long”” of the member function size() to “int” and cast “double” of the floor, ceil, round functions to “int”

lrstat 0.2.72024-06-13

• added the aml, heart, and tobin datasets
• in basket.cpp, replace y[k] = R::rbinom(1, p[j]) with u = R::runif(0,1); y[k] = u < p[j] ? 1 : 0;
• in misc.cpp, replace const in x = NA_REAL with const int x = NA_INTEGER for x = n, n1, y1, n2, y2
• added delete[] for iwork and work pointers in the quad function
• added simonBayesAnalysis and simonBayesSim for Simon’s Bayesian analysis of basket discovery trials
• updated the algorithm for finding the backward image (J, zJ) in getADCI
• updated the initial values for finding the confidence interval in getADCI
• added bmini for optimization and invsympd for inverting symmetric and positive defined matrices to utilities.cpp
• added phregr to estimate hazard ratios from the Cox model for right-censored or counting process data with robust variance
• added liferegr for parametric regression of failure time data with uncensored, right censored, left censored, and interval censored data

lrstat 0.2.62024-05-11

• replaced “R_isnancpp(x)” with “x == NA_INTEGER” for integer parameter x
• replaced the check for equality of spendingTime and informationRates in settings in prints.R

lrstat 0.2.52024-04-22

• updated survQuantile in getDesignMeans.R to handle new censoring and inestimable quantiles
• updated simon2stage to increase the search space of sample size
• added zstatRiskDiff, zstatRiskRatio, zstatOddsRatio, zstatRateDiff, zstatRateRatio for the score test of two-sample crude rates and exposure-adjusted incidence rates
• added mini for minimization of a univariate function over a finite interval
• updated powerRiskDiffExact, samplesizeRiskDiffExact, powerRiskRatioExact, samplesizeRiskRatioExact, powerRiskDiffExactEquiv, samplesizeRiskDiffExactEquiv, powerRiskRatioExactEquiv, and samplesizeRiskRatioExactEquiv to divide the interval of the response rate in the control group into 100 subintervals and obtain the minimum value within each subinterval in order to obtain the global minimum. This replaces the random draw of 500 values of the control group response rate for global minimum. In addition, the attained alpha calculation is updated to use the global maximum
• added riskDiffExactPValue, riskDiffExactCI, riskRatioExactPValue, and riskRatioExactCI for exact unconditional test and confidence limits based on the Miettinen & Nurminen score statistics
• added rawdata to be used as an input data frame to test survival analysis functions
• added kmest and kmdiff for Kaplan-Meier estimates of survival curves and milestone survival difference
• added lrtest for log-rank test using the Fleming-Harrington family of weights
• added rmest and rmdiff for estimates of restricted mean survival times and restricted mean survival time difference
• added lower.tail and log.p arguments to ptpwexp and qtpwexp functions
• added the fquantile function to obtain the quantiles of a survival function
• added the pwexploglik function to obtain the profile log-likelihood function for the change points in the piecewise exponential approximation to a survival function
• added the pwexpcuts function to obtain a piecewise exponential distribution that approximates a survival distribution
• updated getDesignMeanDiffMMRM and getDesignSlopeDiffMMRM to add accrual and dropout information

lrstat 0.2.42024-03-19

• updated the check for pconfigs of getDesignLogistic to avoid an error on macOS
• replaced which_min with which_max in lrsim2e and lrsim2e3a
• replaced qromb with quad for numerical integration
• used numerical integration for lrstat
• renamed kmest to kmstat and use numerical integration
• added kmpower and kmsamplesize for power and sample size calculation for difference in milestone survival probability
• added kmpower1s and kmsamplesize1s for power and sample size calculation for one-sample milestone survival probability
• added rmst, covrmst, and rmstat for restricted mean survival time analysis
• added rmpower and rmsamplesize for power and sample size calculation for difference in restricted mean survival time
• added rmpower1s and rmsamplesize1s for power and sample size calculation for one-sample restricted mean survival time
• updated rmstat, rmpower, rmsamplesize, rmpower1s, and rmsamplesize1s to account for stratification
• added lrpowerequiv and lrsamplesizeequiv for power and sample size calculation for equivalence in hazard ratio
• added kmpowerequiv and kmsamplesizeequiv for power and sample size calculation for equivalence in milestone survival probability difference
• added rmpowerequiv and rmsamplesizeequiv for power and sample size calculation for equivalence in restricted mean survival time difference

lrstat 0.2.32024-02-27

• issued a warning message when unequal spacing is used with O’Brien-Fleming, Pocock, or Wang-Tsiatis boundaries in getBound
• renamed maxInformation to information in the overallResults data frame of the getDesign function output
• added simon2stage for Simon’s two-stage design
• added nbstat to calculate the number of events and information for negative binomial rate ratio
• added nbpower to calculate the power for negative binomial rate ratio test
• added nbpower1s to calculate the power for one-sample negative binomial rate
• added nbpowerequiv to calculate the power for equivalence in negative binomial rate ratio
• added nbsamplesize to calculate the sample size for negative binomial rate ratio
• added nbsamplesize1s to calculate the sample size for one-sample negative binomial rate
• added nbsamplesizeequiv to calculate the sample size for equivalence in negative binomial rate ratio
• added runShinyApp to run a Shiny app for power and sample size calculation for log-rank tests
• added getDesignOneMean for group sequential design for one-sample mean
• added getDesignPairedMeanDiff for group sequential design for paired mean difference
• added getDesignPairedMeanRatio for group sequential design for paired mean ratio
• added getDesignMeanDiff for group sequential design for two-sample mean difference
• added getDesignMeanRatio for group sequential design for two-sample mean ratio
• added getDesignMeanDiffXO for group sequential design for mean difference in 2x2 crossover
• added getDesignMeanRatioXO for group sequential design for mean ratio in 2x2 crossover
• added getDesignPairedMeanDiffEquiv for group sequential design for equivalence in paired mean difference
• added getDesignPairedMeanRatioEquiv for group sequential design for equivalence in paired mean ratio
• added getDesignMeanDiffEquiv for group sequential design for equivalence in two-sample mean difference
• added getDesignMeanRatioEquiv for group sequential design for equivalence in two-sample mean ratio
• added getDesignMeanDiffXOEquiv for group sequential design for equivalence in mean difference in 2x2 crossover
• added getDesignMeanRatioXOEquiv for group sequential design for equivalence in mean ratio in 2x2 crossover
• added getDesignWilcoxon for group sequential design for two-sample Wilcoxon test
• added getDesignMeanDiffMMRM for two-sample mean difference at the last time point from the MMRM model
• added getDesignMeanDiffCarryover for direct treatment effects in crossover trials accounting for the carryover effects
• added getDesignANOVA for one-way analysis of variance
• added getDesignANOVAContrast for one-way analysis of variance contrast
• added getDesignRepeatedANOVA for one-way repeated analysis of variance
• added getDesignRepeatedANOVAContrast for one-way repeated analysis of variance contrast
• added getDesignTwoWayANOVA for two-way analysis of variance
• added getDesignOneSlope for group sequential design for one-sample slope
• added getDesignSlopeDiff for group sequential design for two-sample slope difference
• added getDesignSlopeDiffMMRM for two-sample slope difference from the MMRM model
• added getDesignOneProportion for group sequential design for one-sample proportion
• added getDesignPairedPropMcNemar for group sequential design for McNemar’s test for paired proportions
• added getDesignRiskDiff for group sequential design for two-sample risk difference
• added getDesignRiskDiffExact for exact unconditional test for risk difference
• added getDesignRiskRatio for group sequential design for two-sample risk ratio
• added getDesignRiskRatioFM for the Farrington-Manning score test for risk ratio
• added getDesignRiskRatioExact for exact unconditional test for risk ratio
• added getDesignOddsRatio for group sequential design for two-sample odds ratio
• added getDesignRiskDiffEquiv for group sequential design for equivalence in two-sample risk difference
• added getDesignRiskDiffExactEquiv for exact unconditional test for equivalence in risk difference
• added getDesignRiskRatioEquiv for group sequential design for equivalence in two-sample risk ratio
• added getDesignRiskRatioExactEquiv for exact unconditional test for equivalence in risk ratio
• added getDesignOddsRatioEquiv for group sequential design for equivalence in two-sample odds ratio
• added getDesignFisherExact for Fisher’s exact conditional test for two proportions
• added ClopperPearsonCI for Clopper-Pearson confidence interval for one-sample proportion
• added survQuantile for Brookmeyer-Crowley confidence interval of quantiles of right-censored time-to-event data
• added mTPI2Table for mTPI-2 decision table
• added BOINTable for BOIN decision table
• added mnRiskDiffCI for the Miettinen-Nurminen score confidence interval for two-sample risk difference
• added mnRiskRatioCI for the Miettinen-Nurminen score confidence interval for two-sample risk ratio
• added mnOddsRatioCI for the Miettinen-Nurminen score confidence interval for two-sample odds ratio
• added mnRateDiffCI for the Miettinen-Nurminen score confidence interval for two-sample rate difference
• added mnRateRatioCI for the Miettinen-Nurminen score confidence interval for two-sample rate ratio
• added getDesignOneMultinom for one-sample multinomial response
• added getDesignTwoMultinom for difference in two-sample multinomial response
• added getDesignTwoOrdinal for Wilcoxon test for two-sample ordinal response
• added getDesignOrderedBinom for Cochran-Armitage trend test for ordered multi-sample binomial response
• added getDesignUnorderedBinom for unordered multi-sample binomial response
• added getDesignUnorderedMultinom for unordered multi-sample multinomial response
• added getDesignLogistic for logistic regression
• added getDesignAgreement for Cohen’s kappa agreement coefficient
• added getDesignOneRateExact for exact test of one-sample Poisson rate
• added ptpwexp for distribution function of truncated piecewise exponential distribution
• added rtpwexp for random number generation of truncated piecewise exponential distribution
• added hedgesg for Hedges’ g effect size estimate and confidence interval
• added getDesignEquiv for a generic group sequential equivalence design
• added remlRiskDiff for REML estimates of individual proportions with specified risk difference
• added remlRiskRatio for REML estimates of individual proportions with specified risk ratio
• added remlOddsRatio for REML estimates of individual proportions with specified odds ratio
• added remlRateDiff for REML estimates of individual rates with specified rate difference
• added remlRateRatio for REML estimates of individual rates with specified rate ratio

lrstat 0.2.22023-12-22

• added the intnorm utility function to integrate a function with respect to a normal density
• added predictive power calculation to adaptDesign
• added the ftrunc function to calculate the adjusted p-values for truncated Holm, Hochberg, or Hommel procedures
• reuse the efficacy and futility stopping boundaries calculated under H1 for H0 in lrsamplesize
• added capabilities to calculate Haybittle & Peto boundaries in getDesign, lrpower, and lrsamplesize
• used informationRates as event fractions for conventional log-rank test and information fractions for weighted log-rank tests in lrpower and lrsamplesize
• match the number of events under H0 with the number of events under H1 for conventional log-rank test and match the information under H0 with the information under H1 for weighted log-rank tests
• removed getCriticalValues and getCumAlphaSpent function in lrstat.cpp
• adjusted test-f_lrpower and test-f_lrsamplesize to reflect changes to the definition of informationRates
• renamed informationTime to informationRates in lrsim for consistency

lrstat 0.2.12023-11-28

• used Markdown for Roxygen documentation
• renamed getAccrualDuration to getAccrualDurationFromN
• replaced predictEventOnly with predictTarget for the lrstat function
• added number of subjects reaching the maximum follow-up for fixed follow-up design for the lrstat function
• added efficacyStopping to the getBound function to improve coding efficiency
• added the getPower utility function to improve coding efficiency
• applied only equal spacing of looks for typeAlphaSpending of “OF”, “P”, or “WT” in the getBound function
• replaced the drift parameter with Imax and theta parameters in the getDesign function
• calculated alpha when critical values are not missing for the getDesign and lrpower functions
• added expected information under H0 to the getDesign function output
• added rejectPerStageH0, futilityPerStageH0, cumulativeRejectionH0, cumulativeFutilityH0, and attainedAlpha to the output of the getDesign function
• added the getCI function for parameter estimation after termination of a group sequential trial
• added the getRCI function to calculate repeated confidence intervals of a group sequential trial
• added the adaptDesign function for sample size re-estimation and conditional power calculation
• added the getADCI function for parameter estimation using the backward image method after termination of an adaptive group sequential trial
• added the getADRCI function to calculate repeated confidence intervals for an adaptive group sequential trial
• added the getCP function to calculate the conditional power when the parameter value may vary over time

lrstat 0.2.02023-10-01

• added fadjpdun to calculate the adjusted p-values for Dunnett-based graphical approaches.

lrstat 0.1.152023-08-17

• added fstp2seq for stepwise gatekeeping procedures with or without retesting for multiplicity problems involving two sequences of hypotheses.
• added fstdmix to obtain adjusted p-values for standard mixture gatekeeping procedures
• added fmodmix to obtain adjusted p-values for modified mixture gatekeeping procedures

lrstat 0.1.142023-06-09

• added the getAccrualDuration function to obtain the accrual duration to enroll the target number of subjects.
• added the getDurationFromNevents function to obtain a range of accrual duration to reach the target number of events.
• added the getNeventsFromHazardRatio function to obtain the required number of events given the hazard ratios under the null and alternative hypotheses for a group sequential design.
• allowed studyDuration < accrualDuration + followupTime for fixed follow-up in lrpower
• updated the handling of rounding for fixed follow-up design in lrsamplesize
• updated the handling of null hypothesis for fixed follow-up design

lrstat 0.1.132023-05-20

• added a rounding argument to lrsamplesize to round up the total sample size and events at each stage.
• added by treatment counts of events, counts, and subjects to lrpower output.
• added results under H0 to lrsamplesize output.

lrstat 0.1.122023-04-30

• used tolower to make typeAlphaSpending and typeBetaSpending into case insensitive inputs.

lrstat 0.1.112023-02-09

• added the Kaplan-Meier estimate of milestone survival, Greenwood variance estimate, difference in milestone survival, and Z test statistic for survival difference.

lrstat 0.1.102022-10-25

• added drift parameter to the getDesign function to compute power given the drift parameter.
• updated the repeatedPValue function to respect the range of repeated p-values and to allow matrix input of raw p-values.
• removed repeatedPValueFlag from the fseqbon function.
• removed numSubintervals from the caltime function.
• updated the description of selected functions, parameters, and output.

lrstat 0.1.92022-10-09

• added fwgtmat and fadjpsim to calculate the adjusted p-values for Simes-based graphical approaches.
• updated the print method for design, lrpower, and lrsim.

lrstat 0.1.82022-09-11

• added spendingTime to getDesign, lrpower, and lrsamplesize to allow the error spending time to be different from the information time.
• rewrote lrsamplesize to simplify and accelerate the computation for typeOfComputation == “Schoenfeld”.
• added getBound to obtain the efficacy stopping boundaries for a group sequential design allowing the error spending time to be different from the information time.
• added fadjpbon to obtain the adjusted p-values for graphical approaches using weighted Bonferroni tests for fixed design.
• added updateGraph to update the weights and transition matrix after removing a hypothesis from the set of indices of yet to be rejected null hypotheses.
• added repeatedPValue to Obtain the repeated p-values for a group sequential design based on a given alpha spending function.
• added fseqbon to obtain the test results for group sequential trials using graphical approaches based on weighted Bonferroni tests with the option to provide repeated p-values for each hypothesis over time.
• added lrsim3a to perform simulation for three-arm group sequential trials based on weighted log-rank test. The looks are driven by the total number of events in Arm A and Arm C combined.
• added lrsim2e to perform simulation for two-endpoint two-arm group sequential trials based on weighted log-rank test. The first few looks are driven by the total number of PFS events in two arms combined, and the subsequent looks are driven by the total number of OS events in two arms combined.
• added lrsim2e3a to perform simulation for two-endpoint three-arm group sequential trials based on weighted log-rank test. The first few looks are driven by the total number of PFS events in Arm A and Arm C combined, and the subsequent looks are driven by the total number of OS events in Arm A and Arm C combined.

lrstat 0.1.72022-08-29

• added getDesign for creating a generic group sequential design with constant treatment effect over time.

lrstat 0.1.62022-08-09

• added capability for performing noninferiority tests in lrpower, lrsamplesize, and lrsim.
• added capability for simulating analyses based on calendar times in lrsim.
• adjusted the critical value at the final look if the observed total number of events is less than the planned total number of events in lrsim.
• retained summary statistics for all stages even after crossing the efficacy and futility boundaries in lrsim.
• added number of dropouts to lrpower/lrsamplesize and lrsim output.
• added Schoenfeld method for proportional hazards and conventional log-rank test in lrpower and lrsamplesize.

lrstat 0.1.52022-06-09

• replaced Inf with 6 and -Inf with -6 for test statistic stopping boundaries to avoid potential memory issue.

lrstat 0.1.42022-06-07

New features

• added capability for lrstat to calculate hazard ratios from weighted Cox regression model.
• added capability for lrsamplesize to calculate absolute accrual rate from relative accrual rate given power, accrual duration, and follow-up duration.

Bug fixes

• used specified informationRates to calculate Wang-Tsiatis boundaries.
• used hazard ratios from weighted Cox regression model to determine crossing boundaries on the hazard ratio scale for lrpower.
• replaced stratum-specific output with overall results for lrstat.
• removed hazard ratio estimate from weighted log-rank test from lrsim output.

lrstat 0.1.32022-02-09

• added more statistics to lrpower output.

lrstat 0.1.22022-02-02

New features

• added capability for lrpower and lrsamplesize to use error spending functions.
• added more statistics to lrstat, lrpower and lrsim output.
• allowed user to specify numSubintervals to control approximation.

Bug fixes

• added parameter checking for lrpower, lrsamplesize, and lrsim.
• added test files.
• added print_lrpower.R to print lrpower objects.
• used informationTime instead of informationRates in lrsim to differentiate information based on weighted log-rank tests score statistic variance from information based on number of events.
• renamed sumstat to overview in lrsim output.

lrstat 0.1.12021-12-23

• fixed hyperlinks.

lrstat 0.1.02021-12-17

• Initial release.