Regression Models for Interval Censored Data

Regression models for interval censored data. Currently supports Cox-PH, proportional odds, and accelerated failure time models. Allows for semi and fully parametric models (parametric only for accelerated failure time models) and Bayesian parametric models. Includes functions for easy visual diagnostics of model fits and imputation of censored data.


UPDATES: v2.0.9 Fixed bug when fitting Bayesian model without covariates Changed C++ PI to M_PI at Dirk's request Added ir_clustBoot to account for repeated measures Reduced run time for survCIs example (switched from ic_bayes to ic_par)

v2.0.8 Allows for weights in ic_np Allows plot_legend = F for ic_npList

v2.0.7 Fixed bug with linear predictor being offset pointed out by Ferenci Tamas

v2.0.6 Added JSS citation

v2.0.4 Fixed bug in NAs being mishandled

v2.0.3 Fixed bug with expanding splines Offset covariates to have mean 0 for faster computation

v2.0.2 Fixed bug regarding releveling with a single covariate with only two levels inside formula Internal changes to plotting Allowed p OR q to be supplied to survCIs Allowed MCMC chains to be run in parallel

v2.0.1 Fixed rownames bug for imputeCens when imputeType = ‘median’ Fixed open/closed interval bug pointed out by Dr Bogaerts Added sampling from posterior survival curves Added confidence intervals Added names(fit) method

v2.0.0 Added coef active binding for regression models (i.e. fit$coef is allowed) Fixed issue with fitting model with no regression parameters Reorganized code files Added Bayes models

v1.3.6 Allow for declaration of whether intervals are open or closed for ic_sp and ic_np Added simulationFunctions.R Uses Roxygen for man files

v1.3.5 Added parametric AFT models fixed bug in plot.icenRegFit regarding xlim and ylim for semiparametric model sped up ic_sp added vignette

v1.3.4 fixed show() for summary class with no covariates added ic_np for fitting NPMLE specifically (much faster) switched syntax from “final_llk” to “llk” for all models made plot() and lines() more user-friendly Switched to RcppEigen, rather than including static C++ library

v1.3.3 Fixed memory access issue reported by CRAN

v1.3.2 Preparing for JSS submission Removed unpublished algorithms and methods

v1.3.1 Imported method as reported by CRAN

v1.3.0 getFitEsts is much faster added predict function added imputeCens function added generalized gamma distribution to choice of parametric families added imputed cross validation added plot method for sp_curves numerically stabilized ic_sp(model = 'ph')

v1.2.8 Resolved issue of licenses and gave proper credit to the Eigen team for using their code for matrix algebra Switched to reference classes for fits, rather than lists with reassigned classes Fixed bug with getFitEsts Added methods for calculating the NPMLE (univariate or bivariate) ic_par sped up, especially for models with lots of covariates included use of cbind for response (i.e. cbind(l, u) ~…), rather than Surv(l, u, type = ‘interval’) ~ …, which is still supported Added constrained gradient ascent step to ic_sp Switched to partially numeric derivatives for ic_par, greatly speeding up calculations for models with many covariates

v1.2.7 Added dependencies to “methods” package into NAMESPACE and DESCRIPTION

v1.2.6 models allow for no covariates summary(fit) now returns a summary object rather than just printing At C++ level, switched isnan to ISNAN removed browser() call

v1.2.5 Adding “weights” options to model fits Cleaned up manual Added warning to diag_covar if incorrect yType selected Added essIncData and essIncData_small datasets Switched to analytic derivatives for ic_sp Stabilized ic_par algorithm Preprocessed covariates with PCA

v1.2.1-4 Fixing bugs reported by CRAN check

v1.2.0 Added fully parametric models Added diagnostic tools diag_baseline and diag_covar Switched to Active Set Algorithm for semi-parametric models Added proportional odds model At C++ level, switched to abstract class IC_OptimInfo instead of ICPH_OptimInfo Added simulation of proportional odds

v1.1.1: Stricter convergence criteria (10^-10 instead of 10^-7) Allows for bs_samples = 0 in ic_ph Allows for fitting of models with only 1 covariate Fixed bug in parameterization of imputation model Fixed bug that imputation model sometimes simulates Inf, breaking coxph Added dataset (mdata) slight modification in plot of survival curve

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.


2.0.9 by Clifford Anderson-Bergman, 4 months ago

Browse source code at

Authors: Clifford Anderson-Bergman

Documentation:   PDF Manual  

Task views: Survival Analysis, Missing Data

LGPL (>= 2.0, < 3) license

Imports foreach, methods, MLEcens

Depends on survival, Rcpp, coda

Linking to Rcpp, RcppEigen

Imported by ICcalib, ICcforest, LTRCtrees.

Depended on by icsurvROC.

Suggested by icensBKL.

See at CRAN