Minerva Pappu

Designed and implemented hierarchical Bayesian models for probabilistic reversal-learning behavioral data to estimate subject-level beliefs and volatility. Applied signal detection theory and Bayesian graphical models to analyze distorted-speech and psychiatric symptom data, preparing materials for a grant renewal. Conducted quantitative social network analysis and built/maintained research infrastructure including MongoDB data stores, reusable Python pipelines, and GitHub-based analysis workflows. Scaled Bayesian model fitting on an HPC cluster using Slurm, and migrated research web infrastructure to PHP 8 while keeping data collection running.

Developed a linear-nonlinear neural encoding model to predict neuronal firing rates and spiking activity from electrophysiology recordings. Modeled neural responses using an inhomogeneous Poisson process to connect experimental stimuli to observed spiking behavior. Implemented the workflow in MATLAB to support analysis and model validation.

Applied HYDRA semi-supervised machine learning to identify subtypes of functional network activity patterns across subjects. Performed network-level association analyses between behavioral measures and connectivity across brain region pairs using permutation testing. Implemented analysis pipelines in MATLAB and Python to support statistical evaluation of experimental hypotheses.

Co-authored an R package designed to support collaborative and reproducible data analysis workflows. The project provides tooling to standardize analysis practices and improve sharing across users. Distributed through the CRAN ecosystem with versioned releases.