NEWS.md
In survHC: Higher Criticism Tests for Non-Proportional Hazard Deviations

survHC 0.1.0

First release: an R port of the Python package lifelines-hc.
higher_criticism_test(), berk_jones_test(), fisher_combination_test() and min_p_test() for two-sample right-censored survival data.
event_pvalues() and suspected_deviations() expose the per-interval hypergeometric p-values and the intervals flagged by the Higher Criticism threshold.
plot_pvalue_profile() draws the signed per-interval p-value profile.
Every entry point is an S3 generic and accepts either two samples as vectors or the usual survival formula, Surv(time, status) ~ group, with data, subset and na.action. cbind(time, status) ~ group and time ~ group work without survival attached, the latter treating every observation as an event. The first level of the grouping variable is group A, so alternative = "greater" tests for excess hazard in the second level. The two interfaces are tested to return bit-identical statistics and p-values; they differ only in that the formula method applies na.action to incomplete rows, where the vector interface rejects them.
hc_version selects the Higher Criticism variant: "dj2008" (default), "dj2004" (equation (2) of the paper), "beta" or "star". The names and the default match the hc_version argument of lifelines-hc and the named methods of the Python multitest package (>= 0.2.2); all four agree with it to machine precision.
No dependencies beyond base R, stats and graphics.