On October 27, at the invitation of the Faculty of Business and Management (FBM), Professor Weining Wang delivered an online lecture titled "Learning Network with Focally Sparse Structure". This is part of FBM’s Reading Week activities.

Prof. Steve WB Liu, Dean of FBM gave a welcome speech

Professor Wang undertook a PhD in Economics at the Humboldt University zu Berlin. She is Chair Professor of Financial Econometrics in Department of Economics at University of York. Her interests include financial econometrics, statistics and time series econometrics, and she has published dozens of professional papers in related fields.

Prof. Weining Wang at the lecture

Professor Weining Wang first introduced the properties of regression equations based on panel data, including the estimation of parameters and their basic assumptions. She indicated that parameters of factors will be involved in the network model when studying multiple factors affecting a company's stock price. This study uncovered the network effect with a flexible sparse deviation from a predetermined adjacency matrix, and a double-regularized, high-dimensional generalized method of moments (GMM) framework is proposed to obtain high-quality estimators for the parameters of interest. Professor Wang then introduced the traditional spatial econometric model, including its defects and solutions, leading to the basic model of concern in this study. Then, Professor Wang briefly introduced some related research, theoretical contribution of this research, and the main research content. She indicated that there are some necessary properties of matrix to ensure the effect of estimation. As for the estimation method of the model, Professor Wang pointed out that the first step is to calculate the Regularized Minimum Distance (RMD) estimator, and the second step is to debias by partialling out the effect of the nuisance parameters. Meanwhile, she pointed out the consistency conditions of the estimator. In addition, Professor Wang introduced the inference of estimators, including the linearization method of estimators and the parameter inference through a high-dimensional Gaussian approximation theorem. Subsequently, Professor Wang demonstrated the results of simulation, including single equation model and multiple equation model, and the results show the advantages of the proposed parameter estimation and inference methods. Finally, in the part of empirical study, Professor Wang indicated that the proposed methodology is employed to study the spatial network effect of stock returns. The results show that the network accounting for latent link structure is sufficiently different from the pre-specified one.

The Audiences

Professor Wang's wonderful lecture immersed the audience. In Q&A part, the audience had an in-depth discussion with Professor Wang on the expression of formulas, estimation of parameters and application of models.

Q&A Session