The time series methodology of the Time Series Lab - Score Edition package is well-founded in the academic world and has appeared in highly respected academic journals. For an overview of score-driven publications we refer to the www.gasmodel.com website. The score-driven methodology was developed independently at VU University Amsterdam and Cambridge University. Currently, the knowledge and experience of both universities have been combined and Professor S.J. Koopman (VU Amsterdam) and Professor A.C. Harvey (Cambridge) are part of the Time Series Lab team.More Information
The Time Series Lab - Score Edition software is proprietary software that can be downloaded using this link and is licensed without cost, including commercial purposes. Make sure to read the license agreement before installing the program. Some users / companies require more specialized features of the software. Therefore, Time Series Lab - Score Edition can be made available in a commercial package, fully customized to suit the clients' needs. Please Contact us for more information about customized versions and pricing of the software.More Information
Score-driven models were proposed in their full generality in Creal, Koopman, and Lucas (2013) as developed at the time at VU University Amsterdam. Simultaneously, Andrew Harvey authored the book Dynamic Models for Volatility and Heavy Tails (2013) in which the mechanism to update the parameters over time is the scaled score of the likelihood function. The score-driven model encompasses other well-known models such as the generalized autoregressive conditional heteroskedasticity (GARCH) model, autoregressive conditional duration (ACD) model, and Poisson count models with time-varying mean.More Information
A five step approach to extract signal from time series
Time Series Lab
This is the start-up screen of Time Series Lab - Score Edition. This page can be skipped, next time the pogram starts, by marking the checkbox at the bottom of the page.
Throughout the program, info buttons are placed to provide the user with additional information. The text behind the info buttons is displayed when the user hoovers with the mouse over the button.
This is the core of the program. It holds buttons to the five main steps (Load data, Model setup, Estimation, Graphical Output, and Forecasting). All textual output of the program is displayed in the text window on this screen. The text output can be stored for further processing.
During estimation of the model, the program returns to this output window to monitor the progress of the estimation. Once the estimation has completed the program automatically shows the graphical output window (discussed on a later slide).
Step 1: Load data
After loading the data ('Load data' button on the left), the database headers are displayed in the 'Database' window. By clicking on the database entries, the data is displayed in the graph window. Before going to the next step (green arrow), the user can apply data transformations to the data, e.g. taking first differences.
Summary characteristics are displayed in the top right corner of the graph. Graphical output can be stored via the buttons at the bottom left of the graphical window.
Step 2: Set model details
Here we construct the model by selecting the probability distribution and model components. The dependent variable needs to be selected (top left) before going to the next step. The default model setup is a model with observations coming from the Normal distribution with time-varying level and static variance but many other compents and probability distributions can be selected. This is why our implemented score-driven methodology is so powerfull, it is extremely flexible. The specified model details are summarized in the lower part of the screen.
Step 3: Estimating the model
The model parameters are estimated by maximizing the likelihood, see also the background section. An overview of the parameters is given on this screen. The user can determine starting values for the optimizer and fix parameters to a user determined value. After estimating the model, a parameter report with standard errors and t-stats can be generated by clicking the 'Parameter report' button. The user can always abort the estimation by clicking the 'Abort' button which becomes visible during estimation. During estimation, the textual output window is visible.
Step 4: Graphical output
If the estimation of the model finished successfully, the components of the model can be inspected in the graphical output window. Components can be switched on and off with the checkboxes in the top left. Model components can be saved by clicking the 'Save all components' button. If both location and scale are time-varying, radio buttons become visible to toggle between Location and Scale components. Note that for some probability distributions only one time-varying parameter (location, scale, intensity, probability) is available.
Step 5: Forecasts
The last big step in the analysis of time series is forecasting. Time Series Lab provides a convenient way of plotting and analysing forecasts by allowing the user to dynamically change the forecast window. The loss functions used to asses the quality of the forecasts are updated with each subtraction or expansion of the forecast window.
The forecasts of each component of the model can be analyzed individually or jointly by ticking the check boxes on the top of the page.
Connecting Academia and the Industry
Advances in time series
The Time Series Lab team is constantly looking for ways to improve the software. We do this by continuously working on time series theory and methodology by means of conducting our own research and studying the work of others. If there is something new in the world of time series, we know it. Our Time Series Lab is a controlled environment in which we test time series methodology and new features before making it available in the Time Series Lab software.
A new feature or newly developed methodology may work well on one time series but does not offer advantages for another time series. We make sure that the methodology in Time Series Lab is thoroughly tested and has proven itself before being made available to you.
On the right, you find research output that currently has our attention in the Time Series Laboratory. The listed journal articles are all related to the score-driven methodology that we apply in the Time Series Lab software package.
- Harvey, A.C. and Ito, R. (2020): Modeling time series when some observations are zero, Journal of Econometrics, 214(1), 33-45.
- Gorgi, P., Koopman, S.J. and Lit, R. (2019): The analysis and forecasting of tennis matches by using a high dimensional dynamic model, Journal of the Royal Statistical Society, Series A, 182(4), 1393-1409.
- Blasques, F., Gorgi, P. and Koopman, S.J. (2019): Accelerating score-driven time series models, Journal of Econometrics, 212(2), 359-376.
- Gorgi, P., Koopman, S.J. and Li, M., (2019): Forecasting economic time series using score-driven dynamic models with mixed-data sampling, International Journal of Forecasting, 35(4), 1735-1747.
- Koopman, S.J., Lucas, A., Lit, R. and Opschoor, A. (2018): Dynamic Discrete Copula Models for High Frequency Stock Price Changes, Journal of Applied Econometrics, 33, 966-985.
Close connection with academia
R. Lit, PhD
Rutger Lit is a research fellow of Vrije Universiteit Amsterdam and has a PhD in econometrics specializing in time series analysis. In 2017, he founded Nlitn which is a company that offers consultancy services. Nlitn also offers full data solution packages to suit the data analysis needs of clients. An example is the Time Series Lab software package.Personal website
Professor S.J. Koopman
Siem Jan Koopman is Professor of Econometrics at the Department of Econometrics, Vrije Universiteit Amsterdam. He is also a research fellow at Tinbergen Institute and a long-term Visiting Professor at CREATES, University of Aarhus.
He held positions at London School of Economics and CentER (Tilburg University), and had long-term visits at US Bureau of the Census, European University Institute, and European Central Bank, Financial Research.Personal website
Professor A.C. Harvey
Andrew Harvey is Emeritus Professor of Econometrics in the Faculty of Economics and Politics, University of Cambridge. He was Professor of Econometrics at the London School of Economics before coming to Cambridge in 1996. His most recent book is a 2013 monograph entitled Dynamic Models for Volatility and Heavy Tails.Personal website
Get in touch
Receive updates from Time Series Lab