effective-sample-size

Markov chain Monte Carlo Standard Errors

When more sampling stops helping: a reasoning model can generate a right answer long before it can pick one. The modal and correlation ceilings of test-time scaling, with paper, figures, and code (Bay & Yearick).

Teaching code for the paper After Chat: one question to several AI agent CLIs, who said what, how much agreement is worth, crossing records, safe agent settings. Standard library, no API keys.

Code for reproducing the results in arXiv:2109.01726

A dataset has no single effective sample size: it depends on the question you ask. Lean proofs, analysis scripts and the neff estimator for my paper, The Exceedance Design Effect (arXiv:2608.21262, concept DOI 10.5281/zenodo.21595640).

Custom multi-chain Metropolis-Hastings MCMC sampler in NumPy for Bayesian Logistic Regression on the OASIS Alzheimer cohort. Features log-posterior derivations, log-sum-exp stability, Gelman-Rubin, Blom's Rank-Normalized Split, and Geyer's Bulk/Tail Effective Sample Size (ESS) diagnostics.

Experimental framework for studying covariate shift, chi-squared divergence, importance weighting, effective sample size, and model generalisation.

Label-shift vs concept-drift experiments and a lightweight logit-offset adapter for prevalence shift in binary classification.

Measure error correlation in an LLM judge panel or voter ensemble, report the effective number of independent votes, and correct the majority vote for correlated voters — including the panel shape where it cannot. Closed-form, label-free, numpy only.

Effective sample size computation for Bayesian Estimation of Network Autocorrelation Model