First public release, extracted from a higher-education research pipeline and generalised to any long panel.
Estimation core verified against synthdid: SDID, SC and DiD estimates, weights and untrimmed placebo standard errors match exactly, with and without covariates.
Fixes found during verification: placebo inference for sc and did estimates; fit diagnostics and the trimming rule now use the covariate-adjusted outcome; project_out_covariates() works with arrays; synthetic-control weights are solved exactly (active-set) rather than stopping short of convergence; exact pre-period fits are flagged.