Software

I build open-source tools for research methods, reproducible workflows, and statistical reporting. These projects contribute to my emerging Computational Credibility research program by making computational methods and workflows more transparent and open to evaluation. Development stages and current capabilities vary by package.

mlmr

mlmr is an R package and Shiny app for fitting, understanding, and reporting mixed-effects and multilevel models. It is designed for researchers, instructors, graduate students, and applied analysts who want a guided model-building workflow without giving up transparent R code.

The app uses lme4 under the hood and supports model setup, centering choices, fixed and random effects, interaction terms, APA-style tables, equations, diagnostics, and reproducible exports.

install.packages("pak")
pak::pak("MarcusHarrisUConn/mlmr")
mlmr::run_mlmr()

mcsimr

Monte Carlo simulation design and reporting | Early development

mcsimr is an R package and Shiny workbench for designing, running, and reporting Monte Carlo simulation studies in the social and behavioral sciences. Its main focus is structural equation modeling with lavaan, with OLS regression available as a simpler starting point.

Researchers can vary population parameters, sample sizes, estimators, missing-data mechanisms, and distributional conditions, then examine performance measures such as bias, coverage, power, and convergence. The workflow supports reproducible R code, parallel execution, checkpoints, and Quarto exports with tables, figures, and reporting materials. The package is under active development, and its API may evolve.

install.packages("remotes")
remotes::install_github("MarcusHarrisUConn/mcsimr")
mcsimr::use_mcsimr_app()

DSEMr

Dynamic structural equation modeling | Experimental foundation release

DSEMr is an R package under active development for dynamic structural equation modeling. It aims to support transparent, reproducible workflows for studying time-series relations within people and variation across people, including methodological research and Monte Carlo studies.

The current estimation scope covers Gaussian AR(1) models for a single time series, with optional contemporaneous covariates, and an experimental two-level Gaussian AR(1) model using an R reference engine. The package combines a lavaan-inspired model language with a Rust numerical backend. Broader latent, categorical, and other DSEM model families remain development targets; this release is not a general replacement for Mplus.

Source installation requires R 4.2 or later and a Rust toolchain with Cargo.

install.packages("remotes")
remotes::install_github("MarcusHarrisUConn/DSEMr")