type Vec = readonly number[] | Float64Array;
/**
 * Result of `fTest`: the F statistic, the two-sided p-value, and the degrees of freedom
 * `df1 = n1 - 1` and `df2 = n2 - 1`.
 */
export interface FTestResult {
    statistic: number;
    pValue: number;
    df1: number;
    df2: number;
}
/**
 * Two-sample F-test for equality of variances. `F = s²₁ / s²₂` on sample
 * variances; two-sided p-value `2·min(F_cdf, 1−F_cdf)`.
 */
export declare function fTest(x: Vec, y: Vec): FTestResult;
/**
 * Result of `jarqueBera`: the JB statistic, its p-value, and the sample skewness and
 * excess kurtosis.
 */
export interface JarqueBeraResult {
    statistic: number;
    pValue: number;
    skewness: number;
    kurtosis: number;
}
/**
 * Jarque–Bera test of normality from sample skewness `S` and excess kurtosis `K`:
 * `JB = n/6·(S² + K²/4)`, asymptotically χ²(2). Bridges descriptive stats ↔ tests.
 */
export declare function jarqueBera(x: Vec): JarqueBeraResult;
/**
 * Result of `kruskalWallis`: the tie-corrected H statistic, its p-value, and the degrees
 * of freedom.
 */
export interface KruskalResult {
    statistic: number;
    pValue: number;
    df: number;
}
/**
 * Kruskal–Wallis H-test (nonparametric one-way ANOVA on ranks). Pools all groups,
 * ranks via {@link rankdata}, and corrects for ties; H is asymptotically χ²(k−1).
 */
export declare function kruskalWallis(...groups: Vec[]): KruskalResult;
/**
 * Result of `wilcoxon`: the statistic `min(W+, W-)`, the two-sided p-value, and the
 * normal-approximation z statistic.
 */
export interface WilcoxonResult {
    statistic: number;
    pValue: number;
    zStatistic: number;
}
/**
 * Wilcoxon signed-rank test (paired, or one-sample vs 0). Drops zero differences,
 * ranks |d|, and uses the normal approximation with continuity correction
 * (matching `scipy.stats.wilcoxon(mode='approx', correction=True)`). Statistic is
 * `min(W⁺, W⁻)`.
 */
export declare function wilcoxon(x: Vec, y?: Vec): WilcoxonResult;
/** Result of `fisherExact`: the sample odds ratio and the two-sided p-value. */
export interface FisherExactResult {
    oddsRatio: number;
    pValue: number;
}
/**
 * Fisher's exact test on a 2×2 table `[[a, b], [c, d]]`. Two-sided p-value by the
 * total-probability method (sum of hypergeometric probabilities no greater than the
 * observed table's), matching `scipy.stats.fisher_exact`. `oddsRatio` is the sample
 * ratio `ad/bc` (SciPy reports the conditional-MLE estimate; the p-value matches).
 */
export declare function fisherExact(table: readonly [readonly number[], readonly number[]]): FisherExactResult;
/**
 * CDF of the studentized range distribution with `k` groups and `df` degrees of
 * freedom — `scipy.stats.studentized_range.cdf`. Integrates the range probability
 * against the χ-scaled denominator density (synchronous Simpson + normalCDF/normalPDF).
 */
export declare function studentizedRangeCDF(q: number, k: number, df: number): number;
/** Quantile (inverse CDF) of the studentized range distribution, via bisection. */
export declare function studentizedRangeQuantile(p: number, k: number, df: number): number;
/**
 * One pairwise comparison from `tukeyHSD`.
 *
 * `groups` holds the two group indices. `reject` is true if the q statistic is above the
 * critical value for `alpha`.
 */
export interface TukeyComparison {
    groups: [number, number];
    meanDifference: number;
    qStatistic: number;
    pValue: number;
    reject: boolean;
}
/**
 * Tukey's HSD (honestly significant difference) post-hoc test across `groups`.
 * Pooled within-group variance, Tukey–Kramer standard errors for unbalanced groups;
 * p-values from the studentized range distribution. Mirrors `scipy.stats.tukey_hsd`.
 */
export declare function tukeyHSD(groups: Vec[], alpha?: number): TukeyComparison[];
export {};
//# sourceMappingURL=hypothesis-extra.d.ts.map