Software

skmle

R implementation of the sieve maximum kernel-weighted log-likelihood estimator (SMKLE). It fits transformed hazards models when a time-dependent covariate is observed only intermittently, at times unrelated to the event time. Each covariate measurement is weighted by how far its observation time sits from the time being modelled, instead of carrying the last value forward, and the baseline is estimated by sieves. Core routines run in C++ via Rcpp. Companion package to the paper Kernel meets sieve (JASA, 2025).

🌐 Website · 📄 Paper

PQTR

MATLAB implementation of partial quantile tensor regression, a partial-least-squares-type algorithm for quantile regression with tensor predictors, together with the simulation examples from the paper Partial quantile tensor regression (JASA, 2024).

📄 Paper

tensory

An R package for tensor algebra and for regression with tensor-valued predictors: dense, sparse, symmetric and factorized tensor classes, the CP and Tucker decomposition families, and three regression models relating an array-valued predictor to a scalar outcome. Includes an R implementation of PQTR alongside sparse partial generalized tensor regression and tensor envelope partial least squares. Naming and argument semantics track the MATLAB Tensor Toolbox.

🌐 Website · 📄 PQTR paper

pcdreg

An R package for semiparametric regression of panel count data, the setting where recurrent events are counted between intermittent examination times but exact event times are unknown. It fits a proportional rate model by nonparametric maximum likelihood through an EM algorithm, and a means model by estimating equations, both accommodating time-varying covariates. Companion package to the paper A robust approach for regression analysis of panel count data with time-varying covariates (Bernoulli, 2024).

📄 Paper

SurvSparse

An R package on CRAN for survival analysis with sparse longitudinal covariates, to which I contribute.

📦 CRAN