Crate solow_stats

Expand description

§solow-stats

Statistical diagnostics and hypothesis tests, validated against an authoritative reference implementation.

The crate provides the standard battery of regression-diagnostic and normality tests, simple weighted descriptive statistics, two-sample location tests, autocorrelation tests, and multiple-testing corrections. Every public quantity is cross-validated to tight tolerances against the reference (see tests/reference.rs).

use ndarray::array;
use solow_stats::durbin_watson;

let resid = array![0.1, -0.2, 0.05, 0.15, -0.1];
let dw = durbin_watson(&resid);
assert!((0.0..=4.0).contains(&dw));

Structs§

AnovaRow: One row of an ANOVA table.
AnovaTable: A full ANOVA table.
ContingencyResult: Result of a contingency-table chi-squared test of independence.
CorrelationResult: A correlation coefficient with its two-sided p-value.
DcovTest: Result of the asymptotic distance-covariance (dCov) test of independence.
DescrStatsW: Weighted descriptive statistics for a 1-D sample.
Description: Summary statistics for a numeric sample (the numeric block of the reference Description). All values are computed with the same conventions as the reference: std uses ddof = 1, percentiles use linear interpolation.
DistDependStat: The full battery of distance-dependence statistics for a pair (X, Y).
FTestResult: Result of a linear-restriction F test.
GofResult: A generic (statistic, p-value) result for the tests in this module.
JarqueBera: Output of jarque_bera.
KappaResults: Results of cohens_kappa.
LjungBox: Per-lag output row of acorr_ljungbox.
Mediation: Specification of a linear mediation analysis (no formulas).
MediationResults: Deterministic point estimates from a linear mediation analysis.
MetaResult: Pooled meta-analysis result.
MetaStudy: One study contribution to a meta-analysis.
MultiTestResult: Result of multipletests: rejection flags and adjusted p-values, both in the original input order.
NormalIndPower: Power of a two-sample z-test for independent samples (normal approximation).
NpTestResult: A one-degree-of-freedom test statistic + p-value.
OaxacaBlinder: Blinder-Oaxaca decomposition model.
OnewayResult: Result of a one-way ANOVA.
PoissonResult: Result of test_poisson_2indep.
TTestPower: Power of a one-sample (or paired) t-test.
TTestResult: Result of a t- or z-test: statistic, p-value, and degrees of freedom.
Table: A two-way contingency table of observed counts.
Term: A named model term and the half-open column range [start, stop) it spans in the design matrix.
ThreeFold: Three-fold Blinder-Oaxaca decomposition of the mean gap.
TostResult: Outcome of a TOST equivalence test.
TukeyHsdResult: Result of a pairwise Tukey HSD comparison.
TwoFold: Two-fold (“pooled”) Blinder-Oaxaca decomposition of the mean gap.
VarianceTestResult: Test-statistic + p-value pair returned by variance tests.
WaldResult: Result of a linear-restriction Wald test.

Enums§

Alternative: Alternative hypothesis for a location test.
AnovaType: Type of sum of squares for anova_lm.
Compare: Comparison target for test_poisson_2indep.
ConfintMethod: Confidence-interval method for a binomial proportion.
FleissMethod: Chance-correction convention used by fleiss_kappa.
LeveneCenter: Centering strategy for Levene’s test.
LillieforsDist: Assumed reference distribution for the Lilliefors test.
MetaModel: Pooling model.
MultiTestMethod: Correction method for multipletests.
NearestMethod: Method used by cov_nearest for the correlation-matrix correction.
OnewayUseVar: How heteroscedasticity is treated in anova_oneway.
PoissonMethod: Test statistic / p-value method for test_poisson_2indep.
ResetAug: Augmentation scheme for linear_reset.
TwoFoldType: Weighting scheme for the non-discriminatory coefficient vector β* used by the two-fold (“pooled”) decomposition. Mirrors the reference two_fold_type options.
UseVar: Whether a two-sample test assumes pooled (equal) or unequal variances.

Constants§

PERCENTILES: Percentile levels reported by describe (the reference default).

Functions§

acorr_breusch_godfrey: Breusch–Godfrey Lagrange-multiplier test for residual autocorrelation.
acorr_ljungbox: Ljung–Box test of autocorrelation in the series x.
acorr_lm: Generic Lagrange-multiplier test for autocorrelation (Engle’s acorr_lm).
aggregate_raters: Aggregate a (subject, rater) assignment matrix into category counts.
anderson_darling: Anderson-Darling test with the Stephens (1974) correction for normality (parameters estimated from the data).
anova_generic: One-way ANOVA from summary statistics. Mirrors anova_generic.
anova_lm: ANOVA table for one fitted linear model. Mirrors the single-model anova_lm(model, typ=...).
anova_oneway: One-way ANOVA from raw samples. Mirrors anova_oneway (no trimming).
as_column: Reshape a 1-D sample into an n × 1 matrix (the column-vector form expected by distance_statistics).
bartlett: Bartlett’s test — classical normal-theory χ² test on log-variance ratios.
chi2_contingency: Pearson χ² test of independence on a contingency table (rows × cols). Returns (χ², p, dof, expected).
cohens_kappa: Cohen’s (simple) kappa with variance, confidence interval and zero-test.
compare_f_test: F-test comparison of a restricted (nested) OLS fit against the full fit.
compare_lr_test: Likelihood-ratio comparison of a restricted (nested) OLS fit against the full fit.
corr2cov: Convert a correlation matrix to a covariance matrix given standard deviations std: Rᵢⱼ · σᵢ σⱼ. Mirrors the reference corr2cov.
corr_clipped: Nearest PSD correlation matrix by a single eigenvalue clip plus rescaling so the diagonal is one.
corr_nearest: Nearest PSD correlation matrix by Higham-style alternating projection.
cov2corr: Convert a covariance matrix to a correlation matrix, Cᵢⱼ / (σᵢ σⱼ).
cov2corr_std: Like cov2corr but also returns the per-variable standard deviations σᵢ = sqrt(Cᵢᵢ) (the reference cov2corr(..., return_std=True)).
cov_cluster: One-way cluster-robust covariance.
cov_hac: Newey–West heteroskedasticity- and autocorrelation-consistent (HAC) covariance with Bartlett weights and a window of maxlags lags.
cov_hc0: White’s HC0 heteroskedasticity-consistent covariance: (X'X)⁻¹ X' diag(uᵢ²) X (X'X)⁻¹.
cov_hc1: HC1: HC0 scaled by n / (n − k) (a degrees-of-freedom correction).
cov_hc2: HC2: per-observation scale uᵢ² / (1 − hᵢ) with leverages hᵢ.
cov_hc3: HC3: per-observation scale (uᵢ / (1 − hᵢ))² with leverages hᵢ.
cov_nearest: Nearest PSD covariance matrix, leaving the variances (diagonal) unchanged.
describe: Compute the descriptive statistics of a numeric sample.
distance_correlation: Empirical distance correlation dCor(X, Y) in [0, 1].
distance_covariance: Empirical distance covariance dCov(X, Y).
distance_covariance_test: Asymptotic distance-covariance (dCov) test of independence.
distance_statistics: Compute every distance-dependence statistic for matched samples x and y.
distance_variance: Empirical distance variance of X, dVar(X) = dCov(X, X).
durbin_watson: Durbin–Watson statistic for first-order serial correlation in resid.
f_oneway: Classic one-way ANOVA F-test (equal variances), as in SciPy’s f_oneway.
f_test: F test on the same restriction R θ = r, with df_denom denominator degrees of freedom (typically n - k for OLS).
fleiss_kappa: Fleiss’ or Randolph’s kappa for multi-rater agreement.
fligner: Fligner-Killeen (1976) rank-based variance test.
hat_diag: Leverage (hat-matrix diagonal) hᵢ = xᵢ (X'X)⁻¹ xᵢ'.
het_arch: Engle’s ARCH Lagrange-multiplier test (het_arch).
het_breuschpagan: Breusch–Pagan Lagrange-multiplier test for heteroscedasticity.
het_white: White’s Lagrange-multiplier test for heteroscedasticity.
jarque_bera: Jarque–Bera test of normality.
kendalltau: Kendall τ-b — with ties correction.
kruskal: Kruskal-Wallis one-way ANOVA on ranks. Returns (H, p) where H approximately follows a χ²(k − 1) distribution.
ks_2samp: Two-sample Kolmogorov-Smirnov test.
kstest_normal: Alias for the Lilliefors normality test, matching the reference name kstest_normal. Returns (statistic, pvalue).
levene: Levene’s test / Brown-Forsythe variant.
lilliefors: Lilliefors test of normality with estimated parameters.
linear_reset: Ramsey’s RESET test for neglected nonlinearity.
mannwhitneyu: Mann-Whitney U test — two-sided, normal-approximation p-value with tie correction (Mann-Whitney 1947).
mcnemar: McNemar’s test on a 2 × 2 paired-binary table.
meta_fixed_effect: Fixed-effect meta-analysis.
meta_random_effects: DerSimonian-Laird random-effects meta-analysis.
multipletests: Adjust a set of p-values for multiple testing.
nct_cdf: CDF of the noncentral t-distribution: P(T ≤ t) with df d.o.f. and noncentrality nc.
nct_sf: Survival function 1 − CDF of the noncentral t-distribution.
omni_normtest: D’Agostino–Pearson omnibus test of normality (the reference’s omni_normtest).
pairwise_tukeyhsd: All pairwise Tukey HSD comparisons.
pearsonr: Pearson product-moment correlation.
proportion_confint: (1 − alpha) confidence interval for a binomial proportion.
proportions_ztest: One- or two-sample test for proportions based on the normal approximation.
robust_bse: Robust standard errors: sqrt(diag(cov)) of any covariance matrix.
runs_test: Wald-Wolfowitz runs test for randomness (dichotomised at the median).
shapiro_wilk: Shapiro-Wilk W test of normality (Royston 1992).
spearmanr: Spearman rank correlation.
srange_cdf: CDF of the studentized range with k groups and df degrees of freedom.
srange_ppf: Quantile (inverse CDF) of the studentized range at probability p.
srange_sf: Survival function 1 − CDF of the studentized range.
test_poisson_2indep: Test the equality of two independent Poisson rates.
ttest_ind: Two-sample independent t-test of mean(x1) - mean(x2) == value.
ttost_ind: Two one-sided equivalence test for two independent samples.
variance_inflation_factor: Variance-inflation factor of column exog_idx of the design matrix exog.
wald_test: Wald test on a linear restriction R θ = r.
ztest: One-sample z-test that mean(x1) == value.

Crate solow_stats

Crate solow_stats Copy item path

§solow-stats

Structs§

Enums§

Constants§

Functions§

Crate solow_stats