Grant: Discovery Grants Program - Individual

Recent innovations in technology have generated research questions which cannot be adequately addressed with classical statistical techniques. My area of research is the development of novel techniques for use when classical methods may not be appropriate. For example: a) Spectrophotometers are used to measure the absorption of light at over a 1000 different wavelengths, and investigators have proposed to use these measurements to characterize a sample. Unfortunately, classical multivariate statistical techniques break down in the presence of such high dimensional data. While techniques have been proposed to address this, they often ""overtrain"", that is they underestimate the true errors of misclassification. I propose to study methods which bias estimates towards zero. This will result in classification rules which better estimate the true errors of misclassification. b) Remote image recordings of an area of interest have allowed investigators to collect data in places only previously dreamed of. Unfortunately however, when these recordings are returned to the lab for analysis, the investigators often then ask ""What Now?"". I am proposing to use stochastic processes to extract more information using current sampling techniques, and also a new sampling strategy which will be able to extract more information than current methods about an area of interest. In both cases, the proposed techniques have the potential to improve precision and reduce the costs of evaluating video recordings. c) Automated monitoring devices have been developed which record the movement of objects through an enclosed space. We have developed a new detector which provides more information than previous detectors, including the number, speed, and size of objects. We are now proposing to build statistical models that allow this new detector to be used as a management tool. While this detector and the methods we propose to develop are widely applicable, we are most interested in their potential to provide cost effective monitoring of small animal populations and we are currently seeking collaborators for this venture.