Academic Journal

Tail estimation methods for the number of false positives in high throughput testing

التفاصيل البيبلوغرافية
العنوان: Tail estimation methods for the number of false positives in high throughput testing
المؤلفون: Holger Rootzén, Dmitrii Zholud
المساهمون: The Pennsylvania State University CiteSeerX Archives
المصدر: http://www.math.chalmers.se/%7Erootzen/papers/HighThroughputTesting100303.pdf.
سنة النشر: 2010
المجموعة: CiteSeerX
مصطلحات موضوعية: Multiple testing, tail distribution function, null distribution
الوصف: In high throughput testing – experiments using very many statistical tests but with a real effect expected only in a small proportion of them – extreme tails of distributions are at the center of interest. Hence Extreme Value Statistics, or EVS, provides a natural analysis tool. Typically a considerable amount of preprocessing of data is done before the tests are made. As a consequence often the true null distribution in high throughput testing is different from the theoretical one. Instead a sample from the true null distribution may be available. We provide EVS methods for accurate estimation of the distribution of the number of false positives and for selecting preprocessing methods. Our methods are based on a simple polynomial model for the behavior of the left tail of the distribution of p-values. We provide asymptotics which motivate this model, and exhibit simple estimators of the parameters of the model, and model checking tools, both for independent data and for dependent data. The methods are tried out on two large scale genomics experiments and on a MRI brain scan experiment aimed at finding brain regions involved in dyslexia. A Matlab GUI implementation of the methods may be downloaded from the web.
نوع الوثيقة: text
وصف الملف: application/pdf
اللغة: English
Relation: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.168.3436; http://www.math.chalmers.se/%7Erootzen/papers/HighThroughputTesting100303.pdf
الاتاحة: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.168.3436
http://www.math.chalmers.se/%7Erootzen/papers/HighThroughputTesting100303.pdf
Rights: Metadata may be used without restrictions as long as the oai identifier remains attached to it.
رقم الانضمام: edsbas.3B9FAEEB
قاعدة البيانات: BASE