- Shrinkage estimation linear panel data models that accommodates time varying fixed effects under weak distributional assumptions
- Double and Single Descent in Causal Inference: Outcome Modelling via OLS exhibits double descent, while Synthetic Control exhibits single descent
- Hierarchical shrinkage estimation with trees
- Semiparametric estimation of treatment effects where outcome distributions are thick tailed and therefore hard to model - model the treatment effect function itself
- Tensor PCA for panel data
- CLTs for dependent data using affinity sets (will require several pots of coffee)
- Review of ML for Economic Forecasting
- Stratified designs for Sample selection and treatment assignment to maximise precision subject to budget constraints
- Long term causal inference under persistent confounding using data combination / surrogates
- DML for average partial effect / causal derivatives with a location-scale model for conditional density
- Universal DiD - replace parallel trends with odds ratio equi-confounding
- Time uniform Confidence Sequences (more pots of coffee)