Enhanced False Discovery Rate (EFDR) is a non-parameteric hypothesting testing procedure used for anomaly detection in noisy spatial signals. The approach, published in Shen et al. (2002), is an extension on standard multiple hypothesis testing approaches (for example those employing the Bonferroni correction, or the standard FDR approach) by reducing the number of tests carried out. A reduced number of tests results in an approach which has more power (a lower Type II error), and hence an increase ability to detect anomalies. EFDR has proven to be vastly superior to the Bonferroni correction and standard FDR: the interested reader is referred to Shen et al. (2002) for a detailed study.

The package EFDR contains the required functions to carry out EFDR in a practical setting. It allows for the possibility of a parallel backend (since the computations are relatively intensive), contains basic interpolation methods to grid data which is spatially irregular, and also contains the more standard methods such as detection using the Bonferroni correction, the standard FDR and the Largest Order Statistic method.


For tutorials on how to use this package, please visit the project index HTML page by typing help(package = "EFDR") after installation.


Open issues


Pavlicov√°, M., Santner, T. J., and Cressie, N. "Detecting signals in FMRI data using powerful FDR procedures." Statistics and its interface 1 (2008): 23-32.

Shen, X., Huang, H.-C., and Cressie, N. "Nonparametric hypothesis testing for a spatial signal." Journal of the American Statistical Association 97.460 (2002): 1122-1140.

andrewzm/EFDR documentation built on May 10, 2017, 9:27 p.m.