Quantitating Heterogeneity

研究者研究了两个与流式细胞术数据分析相关的关于混合物分布分析的主题。研究者为混合物复杂性的推断，即混合物中亚群的数量，建立了严格的理论基础。这种理解允许开发新的方法，提供有限样本的置信度声明，比现有的技术更强大，但同时更普遍适用。 第二个主题解决了两个多变量分布的差异的定量表征，基于这些分布的样本。该研究开发了检测两种分布不同的区域的方法，提供了这些区域的科学意义和可解释的描述，并提供了相应差异的置信度声明。基本的方法也适用于其他问题，如疾病病例的空间集群检测。流式细胞仪是世界上使用最广泛的生物医学仪器之一，但缺乏适当的统计方法来分析该仪器产生的数据是充分利用其潜力的障碍。研究者研究了流式细胞术数据分析中出现的两个主题，这些主题对其他应用也很重要。这些数据由不同的亚群组成，研究者开发了严格的方法来确定可以区分多少个亚群，以及如何表征可能具有不同亚群数量的两个样本的差异。这些都是流式细胞术数据分析的重要任务，迄今尚未得到满意的解决。上述主题的统计方法学的进展在生物医学研究和临床环境中具有重要的潜在影响，并且在生物监测，生态学，天文学和医学成像中的空间集群检测中具有重要的应用。

The investigator studies two topics concerning the analysis of mixture distributions which are relevant for the analysis of flow cytometry data. The investigator develops a rigorous theoretical basis for the inference on the mixture complexity, i.e. the number of subpopulations in the mixture. This understanding allows to develop new methodology that provides finite-sample confidence statements, is more powerful than existing techniques, yet at the same time is more generally applicable. The second topic addresses the quantitative characterization of differences in two multivariate distributions, based on samples from these distributions. The research develops methodology to detect regions where the two distributions differ, provides scientifically meaningful and interpretable descriptions of these regions, and provides confidence statements for the corresponding differences. The underlying methodology is also applicable to other problems, such as the detection of spatial clusters of disease cases.Flow cytometry instruments are among the most widely used biomedicalinstruments in the world, but the lack of appropriate statistical methodologyfor analyzing the data generated by this instrument is a roadblockto harnessing its full potential. The investigator studiestwo topics that arise in the analysis of flow cytometry data,but which are also important for other applications.Those data are comprised of various subpopulations, and the investigatordevelops rigorous methodolgy to determine how many such subpopulationscan be distinguished, and how to characterize differences in twosamples that may have different numbers of subpopulations. Theseare important tasks for the anlysis of flow cytometry data which to datehave not been satisfactorily resolved. Advances in the statistical methodologymethodology for the above topics have the potential of a significantimpact in biomedical research and in the clinical setting, and havefurthermore important applications in the detection of spatial clustersin bio-surveillance, ecology, astronomy, and medical imaging.