NEWS.md

In eventPred: Event Prediction

eventPred 0.2.9

• add the generate_plot and interactive_plot parameters to allow users to decide whether to generate plots and if to generate interactive or static plots

eventPred 0.2.8

• adjust the log-likelihood for Cox model for a fair comparison with parametric regression models
• update the package description to add cox event and dropout model parameterization
• add unit tests

eventPred 0.2.7

• use regression and math notation in eventPred-package documentation
• replace dplyr with data.table
• change time intervals to left-closed, right-open for piecewise exponential distributions
• change the algorithm for fitting piecewise exponential regression to that using profile likelihood
• use x to denote the covariates vector or design matrix with intercept
• use q to denote the length of covariates vector or number of columns of the covariates matrix excluding the intercept
• add enrolldf, dffit and text to fitEnrollment output
• add kmdf, dffit and text to fitEvent and fitDropout output
• add target_t to predictEvent and getPrediction
• add a default value to the x parameter in piecewise exponential utility functions
• add cox as a model option for fitEvent and fitDropout, and update predictEvent and getPrediction accordingly

eventPred 0.2.6

• add fix_parameter to predictEnrollment, predictEvent, and getPrediction to allow the parameters to be fixed at the maximum likelihood estimates instead of being drawn from the approximate posterior distributions

• update the use of showplot with respect to the use of showEnrollment, showEvent, showDropout, and showOngoing in predictEvent

eventPred 0.2.5

• rename the components of fitEnrollment output to fit and fit_plot
• restructure the outputs of fitEvent and fitDropout into a list for by-treatment analysis of model fitting and visualization, where each element in the list corresponds to a specific treatment group and has a dedicated sub-list containing two components with one for fit and the other for fit_plot
• update predictEvent.R, getPrediction.R and app.R accordingly to accommodate the new structure of fitEnrollment, fitEvent and fitDropout outputs
• update the output of event_prediction_after_enrollment_completion vignette
• add the condition of (!is.null(event_fit())) for event_fit_ic and (!is.null(dropout_fit())) for dropout_fit_ic in the shiny app
• minor change to the ui layout of the shiny app
• ensure that the randomization date for new patients is after the cutoff date and the event date for ongoing subjects is after the cutoff date

eventPred 0.2.4

• fitEnrollment.R

  • replace round with formatC to retain the zeros after the decimal point

• fitEvent.R

  • parameterize the exponential distribution in terms of log(rate)

  • update the requirement for fitting a piecewise exponential model

  • update the call to the pwexpreg function

  • ensure the sub plots align on the x axis

  • export the sub plots as a list instead of a plotly subplot object

  • replace round with formatC to retain the zeros after the decimal point

• fitDropout.R

  • parameterize the exponential distribution in terms of log(rate)

  • update the requirement for fitting a piecewise exponential model

  • update the call to the pwexpreg function

  • ensure the sub plots align on the x axis

  • export the sub plots as a list instead of a plotly subplot object

  • replace round with formatC to retain the zeros after the decimal point

• predictEnrollment.R

  • add the 'name' parameter to the Plotly traces to ensure proper legends

  • export the sub plots as a list instead of a plotly subplot object

• predictEvent.R

  • parameterize the exponential distribution in terms of log(rate)

  • ensure simulated time >= 1

  • add the 'name' parameter to the Plotly traces to ensure proper legends

  • export the sub plots as a list instead of a plotly subplot object

• getPrediction.R

  • check the input data to ensure all required columns are present

  • check the input data to ensure none of the required columns have missing values

  • add treatment_description to the input data when treatment is present but treatment_description is missing

  • parameterize the exponential distribution in terms of log(rate)

  • obtain event_fit (event_fit_with_covariates) without regard of the existence of event_prior (event_prior_with_covariates)

  • obtain dropout_fit (dropout_fit_with covariates) without regard of the existence of dropout_prior (dropout_prior_with_covariates)

• utilities.R

  • update the pwexpreg function so that its parameters are consistent with other piecewise exponential functions

  • use the Brent method to fit the piecewise exponential regression model with only one interval and no covariates

• launchShinyApp.R

  • newly added to launch the Shiny app for event prediction

• vignettes

  • add event_prediction_at_the_design_stage.Rmd

  • add event_prediction_before_enrollment_completion.Rmd

  • add event_prediction_after_enrollment_completion.Rmd

  • add event_prediction_incorporating_prior_information.Rmd

  • add event_prediction_incorporating_covariates.Rmd

eventPred 0.2.3

• predictEvent.R

  • check to make sure that dropout_fit is not null before simulating dropout times for new and ongoing patients

eventPred 0.2.2

• eventPred-package.R

  • remove import tmvtnsim rtnorm

  • add import purrr list_c map map_dbl

  • add import stats as.formula model.matrix qlnorm rlogis

• utilities.R

  • add pmodavg for the distribution of model averaging of Weibull and log-normal

  • add ppwexp and qpwexp functions for the piecewise exponential distribution

  • add llik_pwexp for the log-likelihood of piecewise exponential regression

  • add pwexpreg for the regression analysis of piecewise exponential distribution

• fitEnrollment.R

  • use the hessian option in optim to remove the optimHess call in fitEnrollment

• fitEvent.R

  • add covariates to the fitEvent function to fit regression models

• fitDropout.R

  • add covariates to the fitDropout function to fit regression models

• predictEvent.R

  • add covariates_event, event_fit_with_covariates, covariates_dropout, dropout_fit_with_covariates

  • fit the event model with covariates if event_fit_with_covariates is not NULL, and fit the event model without covariates otherwise

  • fit the dropout model with covariates if dropout_fit_with_covariates is not NULL, and fit the dropout model without covariates otherwise

  • generate the event time for new patients separately from the event time for ongoing patients

  • generate the dropout time for new patients separately from the dropout time for ongoing patients

  • apply ceiling to the derived time after comparison of generated survivalTime and dropoutTime

• getPrediction.R

  • add covariates_event, event_prior_with_covariates, covariates_dropout, dropout_prior_with_covariates

  • add penalized log-likelihood (posterior) function with covariates for exponential, Weibull, log-logistic, log-normal, and piecewise exponential distributions

  • simplify the algorithm for combining prior distributions across treatments

  • fit event/dropout models with or without covarites depending on the study stage and the presence/absence of covariates_event and covariates_dropout

  • add subject_data to the output

eventPred 0.2.1

• remove the factor attribute of the treatment_description variable

• add pilevel in the output data set for prediction interval level

• replace treatment_label with treatment_description in observed data for enrollment prediction

• update the upper bound of the cutoff reference line in prediction plot

• retain the plots of enroll_fit, event_fit, and dropout_fit in getPrediction output

• add usubjid and treatment_description to the internal data sets

• round the simulated arrivalTime and time so that the time can be interpreted in days

eventPred 0.2.0

• allow the use of treatment labels for by-treatment prediction

• include usubjid in subject-level data sets

• use the quantile method for predicted date if all simulated data sets attain the target number of events

• add log-logistic event model and log-logistic dropout model

• change parameterization of Weibull distribution to be consistent with log-logistic and log-normal distributions in the AFT family

• add AIC to enrollment, event and dropout model fits

• check the required number of events/dropouts for event/dropout model fits

• add "model averaging" and "spline" as additional dropout_model options

• update enroll_fit, event_fit, and dropout_fit for prior incorporation

eventPred 0.1.5

• update design stage prediction with one treatment arm

• allow ongoing subjects with last known date before data cutoff

• update the calculation of ongoing subjects to accommodate ongoing subjects with last known date before data cutoff

• update time for new subjects to start with day 1 and update totalTime calculation for newEvents to remove double count of day 1

• update predictEnrollment to remove calculation of d0, c0, and r0

• add names to event_pred_day

• add nyears and nreps to prediction results

eventPred 0.1.4

• add validity checks for input dataset variables

• update totalTime calculation for observed data

• use method="Nelder-Mead" as the default optimization algorithm for flexsurvspline

• add by-treatment prediction

eventPred 0.1.3

• update the description of internal datasets

• update summarizeObserved to remove adt from adsl

• add Royston and Parmar (2002) spline event model

eventPred 0.1.2

• add mean and variance to prediction output

• update the BIC weight for model averaging

• add more details for model fit parameters

• add day 1 to enrollment plot

• allow prior piecewise Poisson enrollment and piecewise exponential event or dropout models to have additional cut points beyond the observed data range

• update internal data sets

eventPred 0.1.1

• add stage and to_predict information in getPrediction output

• add the cutoff time point to the number of ongoing subjects

• change the default model for dropout to exponential

• require trialsdt in input data set

eventPred 0.1.0

• added the piecewise Poisson model to fitEnrollment and predictEnrollment at the analysis stage

• added number of dropouts

• added number of subjects at risk

• added a data set when the enrollment has completed

• corrected the x-axis title for predictEnrollment and predictEvent

• updated alogrithm to allow one piece piecewise Poisson enrollment model and one piece piecewise exponential time-to-event model

• modified the weight calculation for model averaging to avoid underflow

• used weighted BIC for model averaging

• removed the dropout_model parameter for summarizeObserved

• changed the default number of knots of the b-spline enrollment model to zero

• replaced first and last with slice of dplyr in summarizeObserved

• improved the initial value for the time-decay enrollment model parameters

• added showplot to fitEnrollment, fitEvent and fitDropout

• sped up the calculations of quantiles

• added target_n to predictEnrollment output and target_d to predictEvent output

• removed the cutoff date from ongoing_pred_df before data cutoff

• restricted enrollment model fitting to the last randomization date

• added piecewise exponential dropout model

• use delta method to obtain the variance of model parameters for pooled population

• replace randomization probabilities with treatment allocation within a randomization block

• allow number of subjects to differ among simulated data sets

• remove custom date axis

eventPred 0.0.1

• Initial release

eventPred documentation built on June 10, 2025, 5:14 p.m.