6th CSDA International Conference on
Computational and Financial Econometrics (CFE 2012)
1-3 December 2012, Conference Centre, Oviedo, Spain



The tutorials will take place on Friday the 30th of November 2012 at Salón cultural CajAstur (c/San Francisco, 4 - entrance by c/Mendizabal). View the location by downloading the map of the center of Oviedo. The number of participants to the tutorials is limited and restricted only to those who attend the conference. For further information please contact Ana Colubi.

TUTORIAL 1: 9:00-13:45 (coffee break at 11:00)

Dynamic models for volatility and heavy tails (Tutorial 1.pdf)

Andrew Harvey, Faculty of Economics, University of Cambridge, UK. Email: Contact

This set of lectures will describe a class of nonlinear time series models known as dynamic conditional score (DCS) models. The models are designed to extract a dynamic signal from noisy observations. The signal may be the level of a series or it may be a measure of scale. Changing scale is of considerable importance in financial time series where volatility clustering is an established stylized fact. Generalized autoregressive conditional heteroscedasticity (GARCH) models are widely used to extract the current variance of a series. However, using variance (or rather standard deviation) as a measure of scale may not be appropriate for non-Gaussian (conditional) distributions. This is of some importance, since another established feature of financial returns is that they are characterized by heavy tails.
The dynamic equations in GARCH models are filters. Just as the filters for linear Gaussian level models are linear combinations of past observations, so GARCH filters, because of their Gaussian origins, are usually linear com- binations of past squared observations. The models described here replace the observations or their squares by the score of the conditional distribution. When modeling scale, an exponential link function is employed, as in expo- nential GARCH (EGARCH), thereby ensuring that the filtered scale remains positive. The unifying feature of the models in the proposed class is that the as- ymptotic distribution of the maximum likelihood estimators is established by a single theorem which delivers an explicit analytic expression for the asymptotic covariance matrix of the estimators. The conditions under which the asymptot- ics go through are relatively straightforward to verify. There is no such general theory for GARCH models.
Other properties of the proposed models may be found. These include analytic expressions for moments, autocorrelation functions and multi-step forecasts. The properties, particularly for the volatility models, which employ an exponential link function, are more general than is usually the case. For exam- ple, expressions for unconditional moments, autocorrelations and the conditional moments of multi-step predictive distributions can be found for absolute values of the observations raised to any power. The generality of the approach is further illustrated by consideration of dynamic models for non-negative variables. Such models have been used for mod- eling durations, range and realized volatility in finance. Again the use of an exponential link function combined with a dynamic equation driven by the con- ditional score gives a range of analytic results similar to those obtained with the new class of EGARCH models.
Download here the material of the course Tutorial 1 - Material

TUTORIAL 2: 15:15-20:00 (coffee break at 17:00)

Functional data analysis.

Hans-Georg Müller, Department of Statistics, University of California, Davis, USA. Email: Contact

The tutorial provides an introduction to this area, emphasizing methods, some computational aspects and various applications for the following and related topics: Functional Principal Component Analysis as a central tool; Functional data designs; Principal Analysis by Conditional Expectation (PACE); Functional linear regression for scalar and functional responses; Functional diagnostics; Functional response and dose-response models; Functional methods for longitudinal data; Generalized functional linear models; Functional polynomial models; Functional additive models; Functional gradient models; Empirical dynamics and dynamics learning.
Download here the material of the course Tutorial 2 - Material