Aspects of time series
1. Managing data
Data management is the largest and most time consuming aspect of time series data analysis. By having clear what is available out there for many different needs and formats we can save time. How the data is prepared for description and analysis will depend on the goals of the analysis and the needs that the analysis will satisfy. In this section, you will find links to resources that help with the task.
2. Data description
Once the time series data has been prepared for analysis, we will need to learn its features, what signal predominates, what is the most important aspect of the time series that we should take into account. To that end, we use tools appropriate to the description of time series. This section specializes in surveying those tools.
3. Data modeling
Once we have understood the problem we want to solve, and once we have managed and described the data, we use what has been done to design the model(s) that we plan to use to fit to a training set of the data. This model can help understand the process generating the data better and to forecast or simulate the future.
Forecasting
Forecasting is the art of telling what will happen in the future using information that we have about the past. Depending on the forecasting needs, the forecasting will be done at scale (for many time series at once), or individually for one time series. What we want the forecasting for has determined what we did in the previous 3 steps. We will be able to forecast what we want if we did (1) (2) and (3) well for that.
Simulation
Sometimes we use the fitted model to simulate the future, given that model. This helps us see the possible paths that a given model could follow just by chance. This helps us understand the uncertainty embedded in any forecast.
Software
Some software has a very long history in the analysis of time series. Other not so much, but keeps evolving. Open source is good when we are learning a topic for the first time. This section will feature aspects of time series software.
Probability
Statistical time series assumes that models are stochastic, the observed variables are realization of random variables. Time series signals are there, but in noise subject to the laws of chance.An understanding of introductory probability is important to understand time series modeling, simulation and forecasting.
Interest groups
People are interested in time series in many different ways. Some are concerned with database management, some with modeling, some with the theory of statistical time series, others with approaches to forecasting at scale, and so on. Depth and breadth of time series knowledge does not hurt.
Teaching
The teaching of time series to undergraduate students is not as predominant as graduate teaching. But reality calls for more inclusion of time series analysis in the undergraduate curriculum. This section will keep us informed of others' approaches to time series teaching.
Statistics and Machine Learning
Some basic background in Statistics and Machine Learning is helpful to do time series analysis.
This section will connect to places where that basic background can be obtained.
Time series in action
Citizen science has brought a large number of projects to the attention of many individuals. Sometimes you can get involved in these projects.
This web site is under construction and is gradually being populated with solutions to exercises, links to slides, videos, programs and other resources for the book: Sanchez, J. "Time Series for Data Scientists" published by Cambridge University Press, 2023.
Thank you for visiting
timeseriestime.org
Contact me
If you have any questions on the contents of this web site, you may contact the author of the book, J.Sanchez.