Dina Research School

Workshop: Dataseries, state-space models, and the Kalman filter.

Koldkærgård Landboskole, 18-19 april 2002

Introduction

With modern sensor technologies very large dataseries (e.g. time series) become available. For instance within animal production, data from water and feed consumption, milk production, activity, and automatic weighing are used in some herds. Within plant production, local climate registrations, e.g. temperature and precipitation data as well as data from remote sensing of crops will be used for decision support. Such dataseries call for flexible methods of modelling and efficient methods for computation. This workshop will focus on recent approaches to modelling based on state-space models and computational methods like the Kalman filter and its extensions.

A fundamental characteristic of a state-space model is the presence of an underlying unobserved latent process whose value at each time point influences the observed value in the dataseries. Such latent processes may represent physiological states such as disease, oestrous and feed motivation, or in plant production the presence of disease or pest in the crop. We shall discuss linear Gaussian state-space models but also consider new non-linear and non-Gaussian models.

The Kalman filter is an efficient algorithm for assessment of the values of the latent process and for online updating as new observations become available. This for instance has applications for monitoring where a drastic change in the latent process may call for intervention of some sort.

The workshop will consist of lectures on methodology and case studies with examples of applications of the methodology. The lectures will be accompanied by computer exercises where the workshop participants can get a hands on experience with the methodology.

The workshop is aiming at

• Ph.D.-students in the agricultural and biological sciences whose work involves dataseries of some form and who are interested in learning about modern approaches to modelling and computation for dataseries.
• Ph.D.-students with a background in mathematics, computer science, or engineering who are interested in learning more about the applications in the biological sciences.

The DINA Research School is part of a Nordic cooperation. The workshop language will be English unless all participants have Danish as the first language.

We are looking forward to seeing you at Koldkærgård Landboskole!

Author: phd@dina.kvl.dk. Updated: 28 februar 2002