The hardware and bandwidth for this mirror is donated by dogado GmbH, the Webhosting and Full Service-Cloud Provider. Check out our Wordpress Tutorial.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]dogado.de.

SCMs 0.1.1

Resubmission of 0.1.0, which CRAN returned for changes in February 2026.

Bug fixes

• Covariates observed in only some pre-treatment years are aggregated over the observed years. Previously mean() received the missing values, the feature became NA, and it was dropped without a message, so a specification meant to match on many covariates could keep only one.
• Placebo inference keeps control units whose names contain spaces, hyphens or periods. Previously these units were silently left out of the placebo distribution.
• Two-sided placebo p-values (expected_direction = "two_sided") rank units by absolute effect. Previously the unit with the largest effect received p = 0.
• Outcome models no longer share one mutable synthetic-control fit (the augmented SCM model altered the fit used by the others), and ridge/lasso cross-validation folds are deterministic.
• run_spec_curve_analysis() passes cores to spec_curve(), which previously ran single-threaded unless params contained its own cores entry.
• Parallel workers run the same SCMs code as the calling session (fork clusters on Unix); on Windows, a mismatch with the installed package is an error.

Silent failures are now errors

• A failed placebo run stops with the unit’s name instead of being skipped.
• scdata() names donors that have no pre-treatment data instead of failing with an unrelated error.
• Ridge penalty estimation stops instead of falling back to default values.
• plot_spec_curve() stops when SHAP values do not cover the specification grid.

Changes

• expected_direction defaults to "two_sided" in inference_sc(), inference_placebo(), spec_curve() and run_spec_curve_analysis().
• New curve-level statistic curve_stat = "median_tau" (signed median). One-sided curve-level tests (two_sided = FALSE) require it, because the other statistics are non-negative.
• scest() exposes the inner interior-point solver tolerances (margin.ipop, sigf.ipop, bound.ipop).
• SHAP attribution encodes every level of each categorical specification choice.
• plot_spec_curve() always draws Panel B labels as rich text (richtext_feature_labels is removed) and no longer prints SHAP merge diagnostics.
• Requires ‘CVXR’ 1.9.0 or later, which replaced solve() with psolve(). ‘ECOSolveR’ is now imported directly, because ‘CVXR’ no longer provides the default ECOS solver.

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.
Health stats visible at Monitor.