Added candidate_method = "user" to analyze only user-specified multiplets, with support for variable names and validation of node indices and orders.
Added hyperedge_ids to subset_hypergraph() to select existing results without repeating significance testing.
Updated network-based candidate generation to use the preprocessed correlation matrix and effective sample size.
Updated hierarchical validation to compute and cache required subset intervals. Missing or invalid intervals no longer pass validation.
Fixed alpha_boot so that it controls the BCa confidence level, and corrected Monte Carlo permutation p-values to avoid zero values.
Fixed subset_hypergraph() to update hyperedge lists, incidence matrices, result tables, and counts together, while preserving original test results. Node degree measures now follow the active hyperedges.
Fixed the parallel option and restored the previous future plan after testing, including on error.
Updated function documentation and examples.
Accelerated BCa jackknife calculations by evaluating identical response patterns within each multiplet only once and expanding influence values back to all observations, preserving their original weights. The default regression-based calculation when bootstrap replicates reach the sample size is unchanged.