Introduction to time series analysis

Many data sets we work with are ordered in time and we call them time series (TS). For example, daily exchange rates, quarterly unemployment rate, the weekly number of car accidents, daily temperature, monthly sales of a particular product, hourly number of clicks on a web banner, the weekly number of new clients or visitors of a particular shop. The objective is typically a forecast for the next time period. The temporal order typically implies that the time series possesses “memory”, which we try to capture and using specific time series models.

In this short course, we will get familiar with different types of TS and learn basic tools and methods for characterization, modeling and forecasting of TS. More precisely, we will talk about measuring memory of TS using autocorrelation and partial autocorrelation functions; about stationarity and typical patterns/components in a TS; about the most popular class of linear TS - ARIMA processes - and very shortly about non-linear GARCH models and multivariate TS.

Practical session (registration is required):

• Required background:

a good background in statistics and/or econometric; basic experience in R

• Software:

R (R-Studio or Jupyter) with the following packages: xts, zoo, forecast, ggplot2, rugarch, vars

Registration