Aysun Taseli, Ph.D.
Aysun's research is addressing statistical monitoring, cluster detection, and approximation problems involving a new mixed-risk probability distribution that arises as a convolution of non-identical binomial distributions. Sequential probability ratio tests (SPRT) and resetting SPRT charts have been derived for cases for which the outcome of each Bernoulli event or only the total count is known. Accuracy and detection performance is shown to significantly differ than if homogeneity were assumed.
A second area of focus is on detection of geographical clusters under natural heterogeneity via a new risk-adjusted scan statistic we have developed. Since both the SPRT and scan methods are computationally exhaustive, a third research area investigates the efficient approximation methods based on a normalized cumulant based orthogonal polynomial (Gram-Charlier) expansion and saddle point approximations. Important applications in healthcare include patient mortality, care bundles, hospital readmissions, drug abuse across different patient types. Aysun also serves as our resident statistical expert and mentor.
Education
-
BS, Statistics, Middle East Technical University (1998)
-
MS, Industrial Engineering, Middle East Technical University (2004)
-
PhD, Industrial Engineering, Northeastern University (2011)