To accurately interpret data from biology, pharmacy and health sciences, biostatisticians use mathematical models that represent biological processes and mechanisms. The parameters of these models describe an underlying physical setting and one tries to estimate these parameters so that the employed model justly describes the measured data.

The current algorithms and infrastructure, however, are not up for the difficult task of parameter inference due to the increasing complexity of the models (e.g.: non-linear models). Current algorithms and methods that handle non-linear models are implemented in such a way that they can be considered “black boxes”. These are quite cumbersome in use, let alone reprogrammed by biostatisticians that have little to no knowledge of implementing an algorithm in an efficient way.

There are different methods used for parameter inference, each more complex than the previous one. This gives rise to long execution times, which can be a limiting factor in several cases. Besides long execution times, non-linear models do not converge easily; they suffer from local extrema. Profile likelihood mitigates the latter problem, but it requires that the fitting process is repeated several times. One of the objectives of this doctorate is to enhance existing methods and research new methods to reduce execution time. To accomplish this feat, a GPU will be used for parallel calculations.

Another objective of this doctorate is to try to actively involve the user in the modeling process, by means of an interactive environment and visualizations. The user will gain a deeper understanding of inner workings of the employed non-linear models and the corresponding parameters.

There are two research groups involved in this doctorate, namely the Center for Statistics (CenStat), and the Expertise center for Digital Media (EDM). Both research groups are affiliated with the University of Hasselt. CenStat is renowned for their theoretical as well as applied research. They specialize themselves in bioinformatics, biostatistics, epidemiology, mathematical statistics and public health. EDM are experts in computer graphics, image-based rendering, image processing, and high-performance computing.