Back To Index Previous Article Next Article Full Text Supplement


Statistica Sinica 22 (2012), 1507-1538

doi:http://dx.doi.org/10.5705/ss.2010.257





MODEL SELECTION FOR HIGH-DIMENSIONAL,

MULTI-SEQUENCE CHANGE-POINT PROBLEMS


Nancy R. Zhang and David O. Siegmund


Stanford University


Abstract: Change-point models have been widely applied for segmentation of spatial or time-series data. Some recent applications in genomics motivate multi-sequence change-point models for shared changes across multiple aligned sequences. These applications frequently involve data where the number of change-points can be large. In a previous paper we derived a Bayes Information Criterion (BIC) for determining the number of changes in the mean of a sequence of independent normal observations when the number of change-points $m$ is assumed to remain bounded as the number of observations increases. Here we extend that result to the case where $m$ can increase with the sample size and to simultaneous change-points in multiple sequences. Stochastic terms that enter into the new criteria involve integrals and maxima of two-sided random walks with negative drift. The new criteria are applied to the analysis of DNA copy number data.



Key words and phrases: Change-point detection, DNA copy number, segmentation, model selection.

Back To Index Previous Article Next Article Full Text Supplement