Aller au contenu. | Aller à la navigation

Université
Facultés
Études et Formations
Recherche
Service à la société
International

Advanced Topics in Econometrics: Non-Linear Models - Spring 2019


description of the course

The course is open to PhD students and academics willing to learn new developments in econometrics. It more specifically aims to provide an overview of up-to-date techniques of nonlinear modelling and extreme values modelling. It is design to help prepare students for research in the area of economics and finance. The course starts with a presentation of the platform of High Performance Computing from FWB and a brief overview of the Python language. Each following lecture covers a specific topic.  

Registration compulsory before February 13, 2019 : Registration form

 

Calendar 

Five Wednesdays from 1.30 to 3.30 pm and 4 to 6 pm

 

Course content

Lecture 1: Run your simulations on a cluster: HPC landscape in FWB (F. WAUTELET, UNamur) and overview of the python langage (tba)

High-performance computing (HPC) is nowadays an essential tool for researchers and cluster are an excellent platform for solving a wide range of problems. Typical fields of application for HPC are, among others, numerical simulations of complex systems or the processing of large data. This talk will give an overview of the services provided by the CÉCI. The CÉCI is the 'Consortium des Équipements de Calcul Intensif'; a consortium of high-performance computing centers of the five universities in Fédération Wallonie-Bruxelles (FWB - french speaking part of Belgium)  UCLouvain, ULB, ULiège, UMons, and UNamur. High-performance computing environment at UNamur is provided by the Technological Platform High Performance Computing (PTCI).

Prior to attending this session participants can request a CÉCI account to access the CÉCI supercomputing infrastructure by sending an email to ptci.support@unamur.be.

For further information:

 

Lecture 2: Introduction to Extreme Value Theory (A. KIRILIOUK, UNamur)

Extreme-value theory is the branch of statistics concerned with the characterization of extreme events. Extreme events are encountered in a large variety of fields, such as hydrology, meteorology, finance and insurance. Think for instance of the 2008/2009 financial crisis, which is partly blamed on a mathematical model ("the formula that killed Wall Street"), or of the 1953 North Sea Flood, when the dykes were not sufficiently high to withstand the extreme water height. In classical statistics, it is often the behaviour of the mean or average that is of interest, and the central limit theorem states that a properly normalized sample average will tend to a Gaussian random variable. When interested in extremes rather than in averages, we study the normalized sample maximum, which will tend to a so-called Generalized Extreme Value distribution. Relying on such asymptotic results will allow us to estimate the probability of an event more severe than previously observed: for instance, we could estimate the probability of a loss of more than 50% on a certain financial stock, without previously having encountered such a loss.

This course will take a practical approach, introducing the basics of univariate extreme-value theory through a series of examples. The software environment R will be used to illustrate the concepts step-by-step.

 

 

Lecture 3:  On some time-varying parameter processes and their estimations (A. DUFAYS, Université Laval)

Long financial and macroeconomic time series typically exhibit complex dynamics that cannot be captured by standard stochastic processes such as autoregressive models. Consequently, many flexible non-linear models have been proposed over the last three decades. These models are useful in terms of forecasting and of interpreting the dynamic changes. The course will cover the prominent non-linear models used for fitting time series. It shall firstly focus on univariate models to introduce the different types of non-linearity. In particular, the course will present Change-point, Markov-switching and time-varying parameter models and it will briefly cover their estimations. In a second phase, these approaches will be extended to multivariate time series. To deal with the curse of dimensionality that arises in the multivariate context, standard shrinkage methods will be introduced.

 

Bibliography:

  • J. D. Hamilton, “A new approach to the economic analysis of nonstationary time series and the business cycle,” Econometrica , vol. 57, pp. 357–384, 1989.
  • S. Chib, “Estimation and comparison of multiple change-point models,” Journal of Econometrics , vol. 86, pp. 221–241, 1998.
  • P. Giordani and R. Kohn, “Efficient Bayesian Inference for Multiple Change-Point and Mixture Innovation Models,” Journal of Business and Economic Statistics , vol. 26, no. 1, pp. 66–77, 2008.
  • L. Bauwens, J.-F. Carpantier, and A. Dufays, “Autoregressive moving average infinite hidden markov-switching models,” Journal of Business and Economic Statistics , vol. 35, no. 2, pp. 162–182, 2017.
  • G. E. Primiceri, “Time Varying Structural Vector Autoregressions and Monetary Policy,” Review of Economic Studies , vol. 72, pp. 821–852, 2005.
  • T. Park and G. Casella, “The Bayesian Lasso,” Journal of the American Statistical Association , vol. 103, no. 482, pp. 681–686, 2008.
  • A. Bitto and S. Frühwirth-Schnatter, “Achieving shrinkage in a time-varying parameter model framework,” Journal of Econometrics (forthcoming) , 2018.

 

Lecture 4: Spatial Econometrics (N. DEBARSY, Université de Lille)

In this course, I will first present the cross-sectional and  (static and dynamic) panel spatial specifications. The course will then focus on the interpretation of these models. The final part of the class will be dedicated to the modeling of the endogenous interaction matrices and to methods derived to select the most relevant connectivity matrix(ces) in applied work. 

 

Lecture 5: Econometrics and Machine Learning (E. FLACHAIRE, Université Aix-Marseille)

  • Philosophy and general principle
  • Overfitting and Cross-Validation
  • Decision trees, bagging, random forests and boosting
  • Support vector machine and neural networks 
  • Model misspecification detection

 

More information

Location : University of Namur, Faculty of Economics, Social Sciences and Business Administration • Rempart de la Vierge 8, B-5000 Namur • 4thfloor : Room Camille Joset

Credits policy: The course can grant 3 credits in the doctoral program provided full attendance.

Registration fees: Free

Registration: Compulsory before February 20, 2019.  Registration form

Contact for additional information: