E-Locus - Institutional Repository of the University of Crete

Home Collections Type of Work Post-graduate theses

Post-graduate theses

Current Record: 4796 of 6513

[Add to Basket]

Identifier

000366904

Title

Estimation and control of the False Discovery Rate in Bayesian Network Skeleton Identification, with Application to Biological Data

Alternative Title

Εκτίμηση και έλεγχος του ρυθμού ψευδών ανακαλύψεων (FDR) στην ταυτοποίηση σκελετού Μπευσιανών δικτύων, με εφαρμογή σε βιολογικά δεδομένα

Author

Αρμέν, Αγγελος Πασχάλης

Thesis advisor

Τσαμαρδινός Ιωάννης

Abstract

Bayesian networks are graphical models that represent probabilistic relationships among variables with extensive applications including biological data analysis. In this work, we focus on the problem of estimating and controlling the False Discovery Rate (FDR) in learning the skeleton (set of edges without regard of direction) of a network. We present a unified approach to FDR estimation and control in Bayesian network skeleton identification and experimentally evaluate the performance of the most common FDR estimator in both tasks over several networks and sample sizes. We employ simulated data as well as real flow cytometry measurements of proteins and phospholipids in our evaluation. We demonstrate that estimation in some cases is not conservative and strong control is not achieved, while in other cases estimation is overly conservative. After identifying the possible causes of this lack of accuracy, we evaluate several approaches to deal with them. The results of these evaluations indicate that the goal of accurately estimating and controlling the FDR in all cases using the common FDR estimators may be unrealistic. Thus, we pursue the more realistic goal of accurately estimating and controlling the FDR according to a relaxed definition of false discovery. Our work opens new directions in the utilization of the FDR in learning Bayesian network structure and in estimating structural uncertainty in general.

Language

English

Subject

Bayesian Network

Biological Data Analysis

False Discovery Rate

Structure Learning

Ανάλυση βιολογικών δεδομένων

Εκμάθηση δομής

Μπευσιανα δίκτυα

Ρυθμός ψευδών ανακαλύψεων

Issue date