Time series,

Analysis of time series in practise: identification of trends, model fitting, and forecasting

This training is in R: why?  

The course gives an overview of key concepts related to time series analysis and forecasting. The material will cover both basic techniques used in time series analysis as well as selected topics related to model fitting and diagnostics. The special attention will be given to methods used in trend analysis (including both long-term and seasonal trends) as well as to different approaches used in time series forecasting.

Every concept addressed in the course is illustrated with both examples and hands-on exercises using real time series data.

Thanks to self-paced hands-on exercises the participants of the course will gain practical skills related to time series analysis and forecasting.

The key feature of the course is its practical focus and using free software as the basis. It assures that the participants will gain useful practical skills, and after finishing the course they will be able to analyse single-handedly their own data for his/her purposes without the need to purchase any additional software.

What will you learn?

  • You will learn how to identify and account for regular components in your data (such as long-term trends or seasonal changes).
  • You will familiarize with classical and modern approaches related to time series decomposition.
  • You will know how to correctly perform seasonal adjustment of time series data.
  • You will learn how to select the best time series decomposition method for your data.
  • You will learn how to handle specific data properties such as calendar effects or the presence of outliers.
  • During the course you will gain practical skills in time series analysis and forecasting using computer programmes. We use free R Statistical System R or alternatively: ITSM 2000.
  • If the course is based on statistical system R, you will be provided with R-scripts that will greatly facilitate further work with your own data.

For whom is this training?

Employees of departments of economic analysis, controlling, sales, marketing, asset and liability management, treasury and others who:

  • analyse time series related to economy, finance, labour market, industry, electricity market, sales, etc.
  • want to familiarize with methods used to identify and handle regular components in time series,
  • need to forecast time series that exhibit seasonal variation,
  • are interested in using classical and modern approaches to time series decomposition.

Shortened agenda

  • Introduction to time series analysis
  • Preliminary time series analysis
  • Modeling time series — key concepts and main practical aspects
  • Time series forecasting

Full agenda

  1. Introduction to time series analysis
    • examples of time series data in economics, finance, industry, demography and other areas of application
    • main objectives and main tasks in time series analysis
  2. Preliminary time series analysis
    • basic tools used in time series analysis (autocorrelation function ACF, partial autocorrelation function PACF)
    • preliminary data transformations used in time series analysis (differencing, Box-Cox transformation, normalization)
    • visualization and preliminary analysis for selected real time series data
  3. Modeling time series — key concepts and main practical aspects
    • classical statistical models vs. algorithmic approach
    • stationary and non-stationary time series
    • identification and elimination of trends and seasonal components
    • the decomposition of times series
    • model fitting and checking (model identification, estimation of parameters, diagnostics, statistical tests)
    • selection of the best model – possible strategies and criteria
    • fitting and diagnostics for selected time series – practical examples
  4. Time series forecasting
    • pointwise forecasts and interval forecasts (prediction intervals)
    • forecasting using classical statistical models (AR, MA, ARMA, ARIMA)
    • exponential smoothing – different variants (simple exponential smoothing, Holt and Holt-Winters methods)
    • short-horizon and long-horizon forecasting
    • how to assess and compare forecasting accuracy?
    • constructing pointwise and interval forecasts for selected real time series data. Comparing forecasting accuracy obtained using different methods and selecting the best method.

Try again