Getting Started With Time Series Data

Part summary

One of the goals of this textbook is to equip the reader with the most important practical skills for working with time series data as quickly as possible. As such, unlike most other textbooks, we defer discussion of the theoretical foundations of time series analysis to later.

In the first part of this textbook, we instead cover the special characteristics of wrangling (Chapter 2) and visualizing (Chapter 3) time series data. We then introduce data transformations, time series decomposition, summary statistics, and finally an introduction to forecasting.

This part of the book will also double as an introduction to time series modeling packages in R and Python. In this course, we will primarily make use of the fpp3 package associated with the textbook Hyndman and Athanasopoulos (2018).

library(fpp3)
# help(package="fpp3")

Running the above script, we see that loading fpp3 automatically loads tsibble, tsibbledata, feasts and fable, along with several tidyverse packages. It also comes preloaded with a number of datasets that are useful for pedagogical purposes.

Time series software ecosystem

The software ecosystem for time series analysis in both R and Python is quite extensive. Base R already contains many relevant functions for time series analysis. A curated list of useful time series packages in R can be found via the relevant CRAN task view. ¹

In Python, the statsmodels.tsa package provides most of the basic time series analysis tools. darts is a more recent package that contains machine learning models and provides an scikit-learn compatible interface.

Where to find time series data

The packages fpp3, tsibbledata, astsa come preloaded with several time series datasets, which are discussed in Hyndman and Athanasopoulos (2018) and in Shumway and Stoffer (2000). More datasets can be found on popular platforms such as Kaggle, https://paperswithcode.com/datasets, and https://forecastingdata.org/. Macroeconomic data for countries is often freely available. For instance, data on Singapore can be obtained from https://www.singstat.gov.sg/.

Hyndman, Rob J, and George Athanasopoulos. 2018. Forecasting: Principles and Practice. OTexts.

Shumway, Robert H, and David S Stoffer. 2000. Time Series Analysis and Its Applications. Vol. 3. Springer.

---

1. More generally, CRAN task views are a good first place to look whenever you would like to find a piece of R code to perform a common task.