False Discovery Control for Scan Clustering
This chapter describes and summarizes methods for identifying the presence of clusters in a random field. The approach is based on controlling the fraction of false discoveries and considers a density estimator as the test statistic. A procedure called shaving is adopted for correcting the bias of the density estimator. This type of scanning for cluster identification does not use a window of fixed size; the role of the window size is played by the bandwidth of the kernel estimator. Clusters obtained using different bandwidths are combined in order to increase the detection power of the procedure. In this chapter we stress some more intuitive aspects of these procedures and present some applications.
Keywords and phrasesFalse discovery control false discovery rates scan clustering kernel density estimators
- 8.Perone-Pacifico, M., Genovese, C., Verdinelli, I. and Wasserman, L. (2007). Scan clustering: a false discovery approach. Journal of Multivariate Analysis, 98, 1441–Google Scholar
- 10.van der Laan, M., Dudoit, S. and Pollard, K. (2006). Augmentation procedures for control of the generalized family-wise error rate and tail probabilities for the proportion of false positives. Statistical Applications in Genetics and Molecular Biology, 3.Google Scholar