Format

Send to

Choose Destination
Bioinformatics. 2011 Nov 15;27(22):3129-34. doi: 10.1093/bioinformatics/btr542. Epub 2011 Sep 28.

Lower confidence bounds for prediction accuracy in high dimensions via AROHIL Monte Carlo.

Author information

1
Department of Epidemiology and Biostatistics and Institute of Bioinformatics, University of Georgia, Athens, GA, USA. dobbinke@uga.edu

Abstract

MOTIVATION:

Implementation and development of statistical methods for high-dimensional data often require high-dimensional Monte Carlo simulations. Simulations are used to assess performance, evaluate robustness, and in some cases for implementation of algorithms. But simulation in high dimensions is often very complex, cumbersome and slow. As a result, performance evaluations are often limited, robustness minimally investigated and dissemination impeded by implementation challenges. This article presents a method for converting complex, slow high-dimensional Monte Carlo simulations into simpler, faster lower dimensional simulations.

RESULTS:

We implement the method by converting a previous Monte Carlo algorithm into this novel Monte Carlo, which we call AROHIL Monte Carlo. AROHIL Monte Carlo is shown to exactly or closely match pure Monte Carlo results in a number of examples. It is shown that computing time can be reduced by several orders of magnitude. The confidence bound method implemented using AROHIL outperforms the pure Monte Carlo method. Finally, the utility of the method is shown by application to a number of real microarray datasets.

PMID:
21965820
DOI:
10.1093/bioinformatics/btr542
[Indexed for MEDLINE]

Supplemental Content

Full text links

Icon for Silverchair Information Systems
Loading ...
Support Center