Simulation-based
estimation in econometrics
An
overview with applications to probabilistic discrete choice models and
continuous-time financial models
A Short Course by Eric Zivot,
University of Washington, January 20 - 27, 2004
Eric Zivot is the author of the book "Modelling Financial Time Series
in S-Plus" (which accompanies S+FinMetrics) and a consultant for
Insightful with whom he is currently working on methodology for
simulation-based inference.
Location: ETH Main Building,
Rämistr. 101, Hermann-Weyl-Zimmer, HG G43
Time:
Tuesday, January 20, 13 - 15
Friday, January 23, 10 - 12
Monday, January 26, 10 - 12
Tuesday, January 27, 13 -15
Abstract:
These lectures will give an overview of some recent advances in
simulation-based econometric estimation techniques, focusing on the work
performed under the NSF SBIR grant “Next generation component software
for simulation-based econometric estimation” for Insightful Corporation.
The first set of lectures will concentrate on the use of
simulation-based methods for estimating a wide class of probabilistic
discrete choice models including probit and mixed logit models.
Simulation methods are applied to evaluating high dimensional integrals
that are required to compute choice probabilities. Quasi-Monte Carlo
methods utilizing low discrepancy sequences are employed to improve the
efficiency of the simulation methods. Estimation is performed by
maximizing a simulated log-likelihood function, and some of the
practical issues of doing this are discussed. Examples using the new
module S+discreteChoice will also be presented.
The second set of lectures will focus on the use of simulation-based
methods for estimating continuous-time models in finance. Emphasis will
be placed on continuous-time stochastic volatility models for equities
and interest rates. Continuous-time data are simulated through numerical
solution of stochastic differential equations. Estimation of
continuous-time models is achieved using Gallant and Tauchen’s (“Which
Moments to Match”, Econometric Theory 1996) efficient method of moments.
This method of indirect inference requires fitting a flexible auxiliary
model to observed time series, simulating observations from a proposed
continuous-time model, and matching the data with the model by
minimizing a GMM-type objective function. Examples using the new module
S+SNPEMM will be also be presented.
[Talks in Financial and Insurance
Mathematics] [Finance and
Insurance][Department
of Mathematics][ETH
Zürich]
Please send comments and suggestions concerning this page to Gallus Steiger/Jörg Osterrieder,
email: finance_update@math.ethz.ch.
Last update: January 8, 2004