I describe my research goal as to develop novel econometric and machine learning methods to solve empirical problems of finance and economics.

The methods I proposed during my Ph.D. research aimed to solve two challenges, namely high dimension and the approximation of the unknown functions. For high-dimensional problems, we proposed a method called Specification-Lasso, which can achieve variable-selection and model-selection at the same time. We also extended the power enhanced test in a group manner to strengthen the power of the conventional Wald test when the number of coefficients is diverging. In terms of the approximation of unknown functions, my tools are mainly B-splines and orthogonal series.

I also focus on financial markets where the data are more abundant. I am interested in Asset Pricing, Portfolio Selection, and Volatility. We proposed an asset pricing model constructed by characteristics-based factor loadings. We estimated this model and detected arbitrage characteristics through power enhanced tests. Based on this model, We developed a two-step portfolio selection method. In the first step, we constructed factor-mimicking sub-portfolios through a linear combination of factor loadings. In the second step, we allocated our weights to these sub-portfolios through a dynamic single-index function.