Prerequisites

  1. Basic knowledge of probability theory (more precisely: probability spaces, expectation, variance, covariance, conditional probabilities, Markov's and Hoeffding's inequalities, discrete and continuous real-valued random variables, random vectors, law of large numbers, central limit theorem, multivariate normal distribution).
  2. Basic knowledge of statistics (more precisely: descriptive statistics, point estimation, linear regression, least squares estimation, ML-estimation, statistical testing, confidence sets).
  3. Basic knowledge of linear algebra (more precisely: linear equations, matrix algebra, finite dimensional vector spaces, determinants, positive (semi-)definite matrices, eigenvalues and eigenvectors).
  4. Basic knowledge of calculus (more precisely: limits, (partial) differentiation, integration, calculus of several variables).

Aim of the course
For many practical purposes in statistics the assumption that a given set of observations recorded in time has been generated by independent random variables does not serve as an adequate model of reality. Examples are, for instance, the number of an airline's passenger bookings per week, the daily closing values of stock market indices, monthly car accidents, or the yearly beer consumption in a country. Time series analysis accounts for the dependence in observed data that evolves with time.
This course is an introduction to the mathematical modeling and statistical analysis of time series. It starts with the development of the basic concept of stationary stochastic processes, covers the decomposition of time series into trends, seasonalities and residuals, addresses parametric fitting of ARMA models, non-parametric time series analysis, and forecasting procedures. The course content will be accompanied by an introduction to time series analysis by means of the open-source programming language R.
In addition to classical topics in time series analysis, popular time series models that have been found to be effective at modeling non-linear behavior of time series data will be introduced.
After successfully finishing this course, the students are able to:

  • determine descriptive measures of time series.
  • decompose time series into systematic and non-systematic components.
  • fit linear models to time series data.
    • apply techniques for forecasting future values of a time series.
    • take non-linear models into account.
    • use R functions and packages for analyzing time series data.

The time slot for the lecture will consist to 2/3 of lectures and 1/3 discussion of exercises that have to be prepared by the students as homework.

Lecturers
Annika Betken (UT)