The filtering process would produce an error for a single bimodule. Fixed that.
Filter bimodules with scores lower than the BH multiple testing threshold. But this doesn’t help with scrambled much.
Th new function half_permutation_fdr estimates CBCE’s false discovery rate across different values of alpha using half-permutation.
The function filter_bimodules filters bimodules for overlaps.
Allow singleton bimodules (the KISS principle).
Change the output format. result$comms will be all unqiue communities, result$comms.fil will be the filtered communities. result$summary gives a summary.
Brought back the heursitic search!
Allow covariates directly to be passed to our method.
To correct for covariate removal, added a new parameter n.eff. Removal of each covariate, reduces the degrees of freedom and that must be accounted for in the p-value calculations. This doesn’t matter when the sample size is large, but it does make a difference when it is small.
Use conservative multiple testing correction (BY) by default.
Use deterministic starts by default to avoid surpirses
Randomly choose initial sets to starts with.
Parameter to stop extractions from growing very large.
Use local numbering for the extraction index.
- Added a
NEWS.md file to track changes to the package.