False Discovery Rate Smoothing: A Method for Spatially Dependent Multiple Testing

An empirical-Bayes method for exploiting spatial structure in large multiple-testing problems, presented by Wesley Tansey of The University of Texas at Austin. The method, called false discovery rate smoothing, finds spatially localized regions of significant test statistics and adjusts significance thresholds to control the overall false-discovery rate. It is applied to an fMRI experiment on spatial working memory and its code is publicly available in Python and R.

Use Cases

• Detecting spatially localized signals in fMRI data based on the described application to a spatial working memory experiment.
• Improving power in multiple hypothesis testing problems where test statistics exhibit spatial structure, as described in the method's simulation studies.
• Comparing robustness against existing methods for spatially dependent multiple testing, as the description states it is shown to be far more robust.
• Implementing the FDR smoothing algorithm for non-standard high-dimensional optimization problems using the provided augmented-Lagrangian algorithm.

Strengths

• The method is described as exhibiting state-of-the-art performance in simulation studies.
• Publicly available implementation code is provided in both Python and R.
• The description includes a concrete application to an fMRI experiment, demonstrating biological plausibility.

Limitations

• Row count, column details, and dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The last update date is unknown; freshness is unverified.

Provenance

Source

Wesley Tansey, The University of Texas at Austin

Collection Method

Likely contains simulation data and results from an fMRI experiment, as described in the method presentation.