Update compatability with bgms = 0.1.6.4

[sekulovskin]

Update easybgm for full compatibility with bgms >= 0.1.6.4.

Problem / Motivation

easybgm relied on direct field access to internal bgms objects (e.g., fit$arguments$..., fit$posterior_summary_pairwise$...) and legacy version checks (packageVersion("bgms") < "0.1.6"), which are incompatible with bgms 0.1.6.4. Most importantly, continuous and mixed data were routed to BGGM by default, even though bgms now supports all data types.

Proposed Changes / New Functionality

• bgms is now the default for all data types (continuous, ordinal, binary, mixed, blume-capel)
• bgms extractor functions (extract_arguments(), extract_pairwise_interactions(), extract_rhat(), etc.) are now (hopefully) used consistently throughout the package.
• Rescaled parameters added to output: $partial_correlations, $precision_matrix, $log_odds
• BDgraph now correctly fits GGM (not GCGM) for continuous data when having missing values
• Group comparison for continuous/mixed to require explicit package = "BGGM"
• All legacy packageVersion() checks and deprecated parameter names removed
• All the changes in the analyses are detailed in the help files. So it helps to also read those.

bgms is now the default for all types

fit <- easybgm(data, type = "continuous")

Per-variable type specification (bgms only)

fit <- easybgm(data[, 1:3], type = c("ordinal", "ordinal", "continuous"))

Interpretable scales available in output

fit$partial_correlations # for continuous models
fit$log_odds # for ordinal/binary models

BDgraph now correctly fits GGM instead of GCGM when the data has missing values (see also below)

fit <- easybgm(data, type = "continuous", package = "BDgraph")

Open questions:

• How to properly set the baseline category value for blume-capel variables? Say, I have a mixed model where the last three variables are b-c: How do I set the reference cateogry per variable then (i.e., in cases when I don't want all three variables to have the same value for the reference cateogry)?
• As mentioned above, I now force bgms always when type = "binary". I think this is a reasonable choice.

Files Edited

R/easybgm.R — Default routing to bgms for all types; per-variable type handling; rewritten documentation
R/functions.bgms.R — Replaced all direct field access with extractor functions; added interpretable parameter scales
R/functions.bgmscompare.R — Removed legacy code paths; migrated to extractors
R/easybgm_compare.R — Added validation gate for continuous/mixed without BGGM; rewritten documentation
R/summary.easybgm.R — Removed version checks; added note about available parameter scales
R/summary.easybgm_compare.R — Removed version checks
R/plottingfunctions.bgms.R — Removed version checks and legacy save conditionals
R/plottingfunctions.bgmCompare.R — Same cleanup
R/plottingfunctions.easybgm.R — Removed version checks
R/functions.bdgraph.R — Fixed GGM vs GCGM routing for continuous data
DESCRIPTION — Version bumped to 0.4.0; bgms dependency set to >= 0.1.6.4
tests/testthat/test-easybgm.R — Added tests for new functionality
Testing and Validation
Unit Tests: All 237 tests pass (0 failures, 0 warnings, 0 skips)
Numerical Validation: Verified extractor outputs match expected structure across all model types (GGM, OMRF, mixed MRF) including edge cases (single-variable blocks)
R CMD check: roxygen2::roxygenise() and tools::checkRd() pass cleanly for all .Rd files