NFL Veteran Transition: Causal Inference on Player Performance
Published:
Analyzes NFL player performance data to causally isolate the effects of team transitions from individual skill trajectories. Uses hierarchical mixed-effects models for causal inference and machine learning ensembles for prediction, controlling for confounders like age, position, and team quality.
Key work: Hierarchical mixed-effects model decomposing performance into player-level, team-level, and transition effects · Propensity score matching to construct comparable control groups · Gradient boosting ensemble for predictive performance · Visualization of individual trajectories pre/post transition
Stack: Python, R, mixed-effects models, causal inference, propensity score matching, XGBoost