Statistics is based on applied mathematics and includes experimental design, model and method development as well as validation of models and methods. The statistical paradigm includes model specifications with the help of distributions and validation of these models. Interpretations of the statistical analysis are crucial. Observations play an important role and often the main aim is to understand how data has been generated. Statistics is an instrument that has great potential to describe and explain various phenomena, if it is used properly.
We are giving courses both for SLU's undergraduate and graduate level (visit our course page).
Examples of our research is design of experiments, analysis of variance models, linear, bilinear, mixed linear and generalized linear models, multivariate statistical analysis, high-dimensional statistics, statistical models in epidemiology and environmental statistics.
Statistical models for data with weak scale types
Polycoric and polyserial correlation for ordinal data. Generalized linear models. Mixed models for various types of data. Design of experiments. For more information, please contact Ulf Olsson, Ulf.Olsson@slu.se
generalize the theory around linear models (variance- and regression analysis) and classical multivariate analysis (MANOVA). Bilinear models belong to the curved exponential family whereas linear models (univariate and multivariate) belong to the exponential family. Typical examples of bilinear models are models connected to the analysis of repeated measurements or models which are used in the analysis of spatial-temporal relationships. We are right now working on models where rank restrictions are introduced on parameters describing the mean as well as the dispersion. For more information, please contact Dietrich von Rosen, Dietrich.von.Rosen@slu.se
Small area estimation (SAE)
has become popular to utilize. Often survey studies are used where data is collected from large regions. However, data is mostly insufficient to use when studying small regions. On the other side, if we combine the data with other types of information (often register based) and relevant statistical models it is possible to use, also for the small area, the survey data. For more information, please contact Dietrich von Rosen, Dietrich.von.Rosen@slu.se
High-dimensional statistical analysis
We are interested to study asymptotic distributions when the number of variables, p, and the number of independent observations, n, converge towards a constant, c, i.e. p/n → c, where 0<c. The main aim with the project is to develop high-dimensional statistical analysis by combining knowledge from probability theory with knowledge from multivariate analysis through borrowing and developing tools and techniques from random matrix theory and free probability. Special focus is to derive asymptotic distributions for eigenvalues and anti-eigenvalues when there exist many dependent variables. For more information, please contact Dietrich von Rosen, Dietrich.von.Rosen@slu.se
Extreme-value analysis with applications
Statistical extreme-value analysis is concerned with the study of distributions for extreme observations. Of particular interest is to estimate quantiles of
such distributions, which e.g. in the earth sciences are called return levels. A research field at present is to find suitable models for handling covariates
and time dependency.
For further information, contact Jesper Rydén, Jesper.Ryden@slu.se
We offer statistical advice at all levels and welcome collaborations. For more information visit our consultancy page.
Many of our publications concern applications were statistics has been utilized. A number of publications also appear in leading statistical journals. Link to publications: Publications
Employees at the department have recently written a number of books in statistics. Below you find a sample of some of them.