NEWS.md
score_test(): Score (Lagrange multiplier) test, completing the Wald/LRT/Score trinity. Supports univariate and multivariate cases.complement_test(): NOT operation — negates a test (p → 1-p). Connects to equivalence testing.intersection_test(): AND combinator — rejects when all component tests reject. P-value = max(p_i).union_test(): OR combinator — rejects when any component test rejects. Implemented via De Morgan’s law: NOT(AND(NOT(…))).invert_test(): Test-confidence duality — inverts any test constructor into a confidence set via grid search.lower() and upper(): Generic accessors for confidence set bounds.wald_test() now accepts a vcov matrix for multivariate testing.The combination of complement_test(), intersection_test(), and union_test() forms a Boolean algebra over hypothesis tests. De Morgan’s laws hold by construction: union_test(a, b) is literally defined as complement_test(intersection_test(complement_test(a), complement_test(b))).
z_test(): One-sample z-test, the simplest hypothesis test primitivefisher_combine(): Combine independent p-values using Fisher’s method (demonstrates closure property)adjust_pval(): Multiple testing correction as a higher-order functionconfint() methods for wald_test and z_test (test-CI duality)wald_test() p-value calculation: now correctly uses chi-squared(1) distribution for the squared z-statistichypothesis_test(), lrt(), and wald_test()