df_sigma argument to simKHCE() to allow heavier-tailed measurement errors via a t-distribution.rweibullGF(), which simulates random numbers from a Weibull distribution with gamma frailty.rweibullGF().This release (nearly) finalizes the development of the function simKHCE() and should be the preferred version when using this function.
simKHCE() affecting the piecewise-constant kidney failure event generation with time-dependent predicted true eGFR. The uniform random variable for event time was previously sampled per visit (incorrect); it is now sampled once per subject. Event determination across inter‑visit intervals was adjusted to use cumulative hazard crossing with within‑interval interpolation.simTTE() where some first death events were incorrectly classified as censored.simKHCE() function now uses more extreme default event rates based on observed eGFR to make effects more pronounced. KFRT risk is adjusted at the patient level according to observed eGFR: when observed eGFR is above 30 (mL/min/1.73 m2), the event rate is set to a very low value (10E-7); when observed eGFR is at or below 7 (mL/min/1.73 m2), the event rate is set to a very high value (10E5).The vignette on hierarchical composite endpoints now includes complete documentation of the methodology behind the simKHCE() function.
A simple example has been added to the vignette on win statistics to show how the win ratio splits ties proportionally to the observed wins.
m and hce_type to simTTE() to enable simulation of discrete-time outcomes and creation of hierarchical composite endpoints using the move-down approach.deltaWO(), which computes win odds based on a specified threshold for adhce objects.simKHCE() function related to the implementation of sustained decline.simKHCE() affecting the piecewise-constant kidney failure event generation that uses time-dependent predicted true eGFR between visits, ensuring events are determined correctly over inter-visit intervals.summaryWO.formula() that previously caused errors when GROUP values were used.sigma) in the simKHCE() function has been updated. It now depends on the time-dependent predicted eGFR, hence lower eGFR values result in lower variability.simKHCE() has been revised to prevent events when the most recent eGFR is sufficiently high and to trigger events when the most recent eGFR is below the threshold.two_meas, has been added to the simKHCE() function to enable duplicate eGFR measurements at baseline and/or at the end of follow-up. This implementation was suggested by Amy Shi.summaryWO.adhce() results now include cumulative wins by component.simTTE(), simulates an hce dataset with two correlated outcomes under an illness-death model. It allows population heterogeneity in the first event (which also determines correlation among first events), while the risk of the second event depends on the timing of the first event in the same way across treatment groups.calcWINS() for cases where SE_WP_Type = "unbiased", providing Wilson-type confidence intervals for the win probability, net benefit, and win odds, following the approach of Schüürhuis, Konietschke, and Brunner (2025).regWO(), which previously caused the results to depend on the order of the input dataset. This issue also affected the stratWO() function, since it calls regWO(). A similar issue in the IWP() has also been fixed. The bug was reported by Cyrill Scheidegger.hce() function has been updated for consistency with the as_hce() function. Two new arguments, PADY and AVAL0, have been added. The PADY argument serves a similar purpose as now-deprecated ORD argument. With these updates, hce() can produce outputs of class adhce when the AVAL0 argument is provided.summaryWO.adhce() to provide the summaries by GROUP, as opposed to summaryWO.hce(), which works without grouping by this variable.Details have been added regarding the implementation of the simKHCE() function. The function has been updated to return all time-to-event outcomes for each patient in the ADET dataset.
Examples have been added to the calcWINS() implementation to illustrate the differences between the following formulas for the standard error of the win proportion:
the Bamber-Brunner-Konietschke formula (see Bamber, 1975; Brunner and Konietschke, 2025), Brunner-Munzel test (Brunner and Munzel, 2000) based on the DeLong-Clarke-Pearson (1988) formula, and the Somers (1962) formula.
simADHCE() has been replaced by the all_data = TRUE implementation in simHCE().simHCE() now returns an object of a new class called adhce. This class inherits from the hce class, which itself is a subclass of data.frame. The underlying structure of the returned object remains unchanged. The introduction of the adhce class is intended to clearly distinguish these structured outputs from the more general hce objects. Specifically, an adhce object is an analysis-ready hce object that is derived using multiple time-to-event outcomes and a single continuous (ordinal or score) endpoint.as_hce() has been updated to support additional output flexibility. If the input data includes the variables TRTP, GROUP, AVAL0, and PADY, the function will return an adhce object. In this scenario, even if the AVAL variable is present, it will be recalculated based on the provided data to ensure consistency with the adhce structure. If only the TRTP and AVAL variables are available, as_hce() will return a standard hce object. This enhancement allows users to generate either general or analysis-ready hce objects, depending on the available input variables.regWO() and stratWO() are updated to return the confidence interval for the adjusted and stratified (or adjusted/stratified) win probability as well.regWO().plot() method for hce objects (created by the function as_hce()) is updated to include a fill argument for filling the area above the graph.calcWINS() is updated to include the SE_WP_Type argument with default "biased" (original implementation) and a new "unbiased" implementation of the Bamber-Brunner-Konietschke (see Bamber (1975), Brunner and Konietschke (2025)) standard error for the win proportion.IWP() is added to calculated patient-level individual win proportions.as_hce() is added.simHCE() is updated to correct for the copula implementation so that theta = 1 (case of independence) and theta close to 1 now give similar results (as expected).simHCE() is updated to include a new theta argument for Gumbel dependence coefficient of the Weibull distributions for time-to-event outcomes. Default is theta = 1 which assumes independence of time-to-event outcomes. The argument is still experimental.calcWO() is updated to return the confidence interval for the win probability as well.plot() method for hce objects (created by the function as_hce()) is implemented to provide the ordinal dominance graph as suggested by Bamber (1975).powerWO(), sizeWO(), minWO() are updated to include a new argument alternative to specify the class of alternative hypothesis. All formulas are based on the Bamber (1975) paper.COVID19plus.NEWS.md file to track changes to the package.HCE1 - HCE4 datasest are updated to follow the standard structure.dec is added to simHCE() for decimal places used for rounding the continuous outcome in the simulated dataset. Additionally, the default value for the standard deviation of the continuous variable in the placebo group CSD_P is changed to be equal to that of the active group CSD_A instead of being equal to 1.simADHCE() which simulates adhce objects, that is, an hce object with its source datasets. Works similar to simHCE() which provides only an hce object.simORD() which simulates ordinal endpoint by categorizing beta distributions.