Reduced Form Analysis of World Productivity Growth: A Model Averaging Approach.
Formalized both structural model and dynamic reduced form stochastic frontier methods to measure productivity and efficiency;
Introduced model averaging approach into stochastic frontier setting to reduce model uncertainty risk and better approximate underlying data generating process;
Applied Jackknife model averaging criterion to obtain optimal weight for each candidate model and to asymptotically achieve lowest prediction squared errors; implement in Matlab.
Pricing Characteristics: An Application of Shephard’s Dual Lemma
Utilized translog input distance function to model the value generating process in housing market; used corrected OLS, time-dummy LS, and stochastic frontier analysis in estimation; implement in Stata;
Applied Shepard’s dual lemma to calculate shadow prices of property’s characteristics, and constructed Fisher price index for each characteristic;
Shadow prices of all characteristics can be simultaneously in contrast to the hedonic regression of housing market analysis, which needs exogenous assumptions to overcome multicollinearity problem.
Panel Data and Productivity Measurement: An Analysis of Asian Productivity Trends.
Reviewed productivity measurement and discussed empirical technique to decompose total factor productivity (TFP) into innovation and catch-up effects;
Utilized six stochastic frontier models to analyze technical efficiency change and TFP growth rate in Asian countries; implement in Matlab.
Course Project about movie rating prediction and recommendation
Used five methods to predict movie ratings with highly scarce data, including standard SVD, regularized SVD, KNN, and ensemble technique;
Adjusted different tuning parameters to reduce number of steps to convergence and increase prediction precision; achieved prediction RMSE 0.8658.
Projects about stocks selection and performance evaluation
Used three criteria in stock selection, aiming to find stocks that outperform the market.
Constructed back-tests in Matlab to compare performance among groups and with market indices.
