Principal Investigator Roy Welsch
Biological data analyses can be severely degraded by "bad" data points that result from experimental or measurement errors or by small fractions of "good" data that prove to be overly influential and skew the results. Typically, these data points are difficult or impossible to detect a priori. The robust and high-breakdown statistical methods that have been designed to address such problems for other applications are too computationally intensive for high-volume applications such as automated optical imaging. However, these methods can be optimized for specialized applications that require intensive computational analysis.