Projects (Code in Python)
Cross-Validated High-dimensional Conditional Density Estimation [link]
-
A unified framework for estimating conditional density with high-dimensional covariates
- new representation: conditional density
~
many conditional means - allowing for any machine learners of conditional means (e.g
sklearn
) - The estimator is fully data-driven, achieved through cross-validation
- new metric/loss for cross-validation, easy to implement
- theoretical guarantee for the optimality
- new representation: conditional density
Difference-in-Differences Models with Continuous Treatment [link]
-
Extending the diff-in-diff framework to continuous treatment
- estimating average treatment effect on treated (ATT) at any continuous treatment intensity
- under double/debiased machine learning (DML) framework: debiased score + crossfitting
- can accommodate high-dimensional covariates
- the estimator is asymptotically normal with explicit asymptotic variance
- bonus multiplier bootstrap confidence interval
Approximate Sparsity Class and Minimax Estimation [link]
-
Proposing a new type of sparisty: approximate sparsity
- complexity (metric entropy) and minimax rates are established
- LASSO (as a selection mechanism) is still optimal
- data-driven LASSO threshold based on a maximal (Talagrand’s) inequality
- simple code that implements the theoretical results
- post-processing algorithm and additional Monte-Carlo simulations are provided
Consumer Preferences, Choices, and Counterfactuals [link]
-
Implementation of Bayer, Ferreira, and McMillan (2007) in Python
- combination of BLP and intrumental variable
- contraction mapping
- counterfactual estimation
Introduction to Econometrics [link]
-
Jupyter Notebook for undergraduate econometrics course at UCLA
- basic data manipulations (
pandas
andnumpy
) - linear regressions and hypothesis testing (
statsmodels
) - basic plotting (
matplotlib
) - special topic(s) (diff-in-diff)
- basic data manipulations (