TILS619 Time Series Analysis (2–5 cr)
Description
Time-series is a dataset consisting of observations recorded at consecutive time instants. Time-series typically have serial correlation.
We discuss in the course, for instance, the following topics:
- graphical exploration of time-series (for instance using autocorrelation and partial autocorrelation)
- stationarity,
- linear time-series models (ARIMA), choosing a model, estimation of its parameters, and prediction using the model.
The main emphasis of the course is on ARIMA-models, but different instances of the course have varying additional content, such as:
- frequency domain methods (such as periodogram),
- multivariate time-series, and/or
- state-space models
Learning outcomes
Student who completes the course successfully:
- can draw and interpret the sample autocorrelation of a time-series,
- recognises time-series, which can be considered stationary,
- can choose a suitable ARIMA-model for a time-series,
- can investigate the fit of the model
- can use the model for prediction, including a confidence interval.
Additional information
Degree students in statistics and data science take the 5 credit course. Degree students from other fields can take either 2 credit or 5 credit course.
The course will be lectured (at least) every 1.5 years.
Description of prerequisites
,5 credits: Statistical inference 1 and 2; Further probability.
2 credits: Data visualization and analysis and From data to model or equivalent background; familiarity with basic usage of R.