20236 - TIME SERIES ANALYSIS OF ECONOMIC-FINANCIAL DATA
Course taught in English
Go to class group/s: 31
Class 31: SONIA PETRONE
Basic notions of Statistics and Probability.
The analysis of dynamic phenomena is crucially important in economic and financial studies. The course aims at providing solid methodological background and data-analysis skills for time series analysis, covering classical as well as modern techniques for non stationary time series, based on state-space models.
- Aims of time series analysis and descriptive techniques:
- Time series decomposition. Exponential smoothing.
- Probabilistic models for time series analysis:
- Time series as a discrete time stochastic process.
- Stationary processes. Summaries. Estimation of the autocorrelation function.
- First examples: White noise. Gaussian processes. Random walks.
- Categorical time series: Markov chains. Inference for Markov processes.
- Stationary time series: ARMA models (brief review).
- Time series with structural breaks: Hidden Markov Models.
- State space models for time series analysis:
- Motivating examples: non-stationary series; stochastic volatility; streaming data.
- State space models: definition and main properties.
- Hidden Markov models as state-space models.
- Dynamic linear models (DLM).
- Filtering, forecasting, smoothing: Kalman filter and Kalman smoother.
- Innovation process and model checking.
- Maximum likelihood estimation of unknown parameters.
- Examples for economic and financial time series. DLMs for trend, seasonality, cycle.
- Nonlinear regression by DLMs.
- ARMA models as DLMs.
- Multivariate time series (dynamic regression (example: term structure of interest rates); seemingly unrelated time series models; factor models).
- Bayesian inference and forecasting via Markov Chain Monte Carlo (MCMC).
- Recent developments.
- Explain and describe the main statistical methods for time series analysis.
- Identify the models suitable for the problems under study; estimate and make forecasts for dynamic systems, both stationary and non-stationary, with an adeguate quantification of uncertainty and risk.
- Use R for time series analysis.
- Apply and properly interpret the models and methods presented in the course in applications.
- Use adeguate statistical software (R and main R functions for time series analysis).
- Evaluate and justify their analysis on real data.
- Prepare appropriate reports of their statistical analysis in real data applications.
- Face-to-face lectures
- Exercises (exercises, database, software etc.)
- Group assignments
- Exercises: lectures in the computer room ('laboratories') on the analysis of real data. Software: R, freely available at www.r-project.org. An R-package, 'dlm', has been developed for this course.
- Students are involved in the learning process through individual and team work in periodic assignments.
|Continuous assessment||Partial exams||General exam|
- There are no partial exams, but there are about 4 take-home assignments (individual or team work). Assigments are not mandatory, but strongly encouraged for an active learning. They are not evaluated for the final exam; yet, students who did not deliver the assignments have to answer additional questions on data-analysis with R in the written proof.
- A final project on real data analysis (individual or team work) is mandatory and evaluated for the final exam (30%).
- Written proof (70%; it can be 100% if poorly done) .
- C. CHATFIELD, The Analysis of Time Series, Chapman & Hall/CRC, 2004, 6th edition.
- G. PETRIS, S. PETRONE, P. CAMPAGNOLI, Dynamic Linear Models with R, Springer, New York, 2009.
- S. PETRONE, Lecture notes: Introduction to Markov Chains, 2015.
- Lecture notes, data sets, R code, R Markdown templates etc are made available on the Bboard of the course.