Changelog • cpmr

cpmr (development version)

Removed the legacy cpm() entry point and the deprecated confounds alias. Use fit(cpm_spec(...), ...) / fit_resamples(cpm_spec(...), ...) for the native workflow.
fit() now performs a single fit; resampling is handled by fit_resamples().
cpm_resamples no longer stores redundant resample metrics, and the collect_*() helpers have been removed. Use summary() for aggregated resample results and predictions, edges, and folds for raw outputs.

Added cpm_spec() as the native interface for fit() and fit_resamples(), and made single-fit and resample result objects more consistent.
Added summary.cpm_resamples(), which now reports pooled out-of-fold error metrics by default and keeps pooled / fold-wise correlations as supplementary statistics.
Added resample_metrics() for direct access to pooled or fold-wise metric tables from a cpm_resamples object.
Added native-first documentation, including a getting-started vignette and reorganized pkgdown reference pages.

Refactored internals and expanded test coverage around fit, summary, edge, and resampling behavior.
Hardened resample validation so each fold must retain at least 3 complete-case training observations.

CRAN release: 2026-03-11

Renamed internal constant networks and includes to corr_types and inc_edges, respectively, to better reflect their purpose. In addition, the documentation which mentioned networks has been updated to use correlation types to better reflect their meaning.
Reorganized R/cpm.R so cpm() remains the entry-point function at the top, followed by print.cpm(), with internal helpers grouped below.
Improved test architecture with explicit internal-helper coverage and complete-case fold invariants.

Refactored the internal cpm() workflow into smaller helpers (normalize_inputs, resolve_include_cases, apply_confounds_regression, etc.) while preserving user-facing behavior.
Removed the obsolete internal alias regress_counfounds().
Removed broad Rfast namespace import in favor of explicit Rfast:: calls.
Updated summary.cpm() edge summarization path to avoid mutating local object fields.

CRAN release: 2024-10-06

Added summary() method to summarize the results of the CPM analysis (#8).
Added tidy() method to tidy the results of the CPM analysis (#10).
Support na_action argument in cpm() function to handle missing values in the input data (#2).

Added params to cpm() output to store the input arguments (#14).
Let "sum" be the default value for return_edges argument.
Let the first two dimensions of edges in the output be edges and networks, respectively.
Polish the print method of the cpm class.

CRAN release: 2024-06-08

CRAN release: 2024-03-14

Added return_edges argument to optionally set how to return edges in the output.

Fixed confounds treatment. Now confounds are used in feature selection but not in model fitting.