The ORSTAT group has expertise in Operations Research, Statistics and Econometrics. This broad range of skills is reflected in the following description of the research topics of each of the group's members.
 |
 |
 |
 |
 |
|
G. Claeskens |
C. Croux |
R. Leus |
F. Spieksma |
M. Vandebroek |
Gerda Claeskens performs research in model selection, nonparametric estimation and nonparametric testing methods. Some examples of nonparametric models are penalized spline and wavelet models. Recently studied lack-of-fit and goodness-of-fit tests use some form of order selection methods or are Bayesian inspired. Another big component of her research comprises model selection methods (in low and high dimensions) and a study of inference after model selection, and in particular (frequentist) model averaging estimators. This leads to the construction of the new class of focused information criteria.
Christophe Croux is interested in robust methods in statistics, applied and theoretical econometrics, computational statistics, exploratory data analysis and mathematical statistics. Currently, he studies robust discriminant and classification analysis, the econometrics of causality tests in a time series context, variable selection problems for high-dimensional problems. He is also interested in the use of advanced statistical techniques in marketing applications. Application and implementation of the proposed methodology forms an essential part of the research projects.
Roel Leus performs research in operations management and operations research. Specific areas of study are machine scheduling and (resource-constrained) project planning and scheduling, both deterministic and with uncertainty (robust and reactive/dynamic planning). The encountered optimization problems are tackled either by means of exact solution procedures (e.g. branch-and-bound) or using heuristics. Another topic of interest is project portfolio management: the accept/reject decision (selection) of new projects that arise as opportunities to an organization, as well as the prioritization of admitted projects. In this work, it is attempted to take into consideration both capacity (available resources) and the time-dependent value of the project (lead time is inversely related to utilization of the resources), as well as the financial value of the portfolio (as reflected by e.g. the total expected net present value (NPV) of the projects).
Frits Spieksma is interested in combinatorial optimization. Formulating, classifying, and solving combinatorial optimization problems is a challenging research topic. Exact methods, such as branch-and-bound, branch-and-price, are investigated for specific combinatorial optimization problems such as assignment and scheduling problems. Since these methods may be time-consuming, studying heuristic methods is relevant as well. For the latter class of methods, the emphasis is on approximation algorithms that have a provable guarantee with respect to the value found. The problem of partitioning a partial order has also been studied. A winner determination problem in a specific combinatorial auction is also investigated. Research on client-oriented vehicle routing problems is ongoing.
Martina Vandebroek's main research topic is the optimal design of experiments, both in an industrial and marketing context. In industrial experimentation, issues that are dealt with are blocking, split-plotting, trend effects and model misspecification. Currently she focuses on conjoint experiments that are frequently used in marketing but also in transportation, health economics, etcetera. Rating based experiments, in which the respondents rate the different alternatives, ranking based experiments, in which the alternatives are ranked from most to least preferred, as well as discrete choice experiments, in which the respondents choose their preferred alternative in several choice sets, are optimized. Various related aspects are studied: inclusion of a no-choice option, measuring willingness-to-pay, market heterogeneity, choice complexity and online experimentation. Also the related segmentation and computational problems are under investigation.
