Instrumental Variable Analysis in Clinical Research
A practical guide to instrumental variable analysis in clinical research, including confounding by indication, randomized encouragement designs, treatment nonadherence, the IV assumptions, Wald estimators, two-stage least squares, local average treatment effects, diagnostics, R implementation, and a complete worked example.
Generalized Linear Models (GLMs) Overview
A practical clinical biostatistics guide to generalized linear models, covering the exponential family, link functions, logistic regression, Poisson and negative binomial models, model interpretation, estimation, diagnostics, and clinical trial applications.
Checking Regression Assumptions and Diagnostics
A practical guide to checking regression assumptions and diagnostics in clinical research, including linearity, normality, homoscedasticity, independence, influential observations, multicollinearity, residual diagnostics, transformations, and R and SAS examples.
Propensity Score Matching in Observational Studies
A deep, computationally grounded guide to propensity score matching: a real confounding-by-indication demonstration that reverses a truly protective effect, propensity score estimation, matching with real balance diagnostics, and a comparison against covariate adjustment and inverse probability weighting.
Polynomial and Spline Regression for Dose-Response
A computationally grounded guide to polynomial and spline regression for dose-response analysis, including linear, quadratic, cubic, and restricted cubic spline models, knot selection, model comparison, nonlinear effects, and clinical interpretation.
Ordinal Logistic Regression for Clinical Outcomes
A computational guide to ordinal logistic regression and the proportional odds model: cumulative logits, odds-ratio interpretation, predicted probabilities, model assumptions, and what to do when proportional odds fails.
Multiple Linear Regression in Clinical Research
A deep, computationally grounded guide to multiple linear regression: OLS derived in matrix form, a real fitted trial showing the precision gain from ANCOVA-style baseline adjustment, real residual diagnostics, and a genuine multicollinearity failure with exploding VIFs.
Generalized Estimating Equations (GEE) Explained
A deep, computationally grounded guide to GEE: Fisher scoring and the sandwich estimator derived and implemented from scratch, validated against production software, with a real demonstration of population-averaged vs. subject-specific coefficient divergence and GEE