/**
 * Multiple comparison correction for controlling family-wise error rate
 */
/**
 * Bonferroni correction for multiple comparisons.
 *
 * Controls family-wise error rate (FWER) by dividing alpha by the
 * number of tests. Very conservative but simple.
 *
 * @param pValues - Array of p-values from multiple tests
 * @param alpha - Original significance level (default: 0.05)
 * @returns Corrected alpha threshold and which tests are significant
 *
 * @example
 * ```typescript
 * const pValues = [0.001, 0.02, 0.045, 0.08, 0.12];
 * const result = bonferroniCorrection(pValues, 0.05);
 * console.log(result.correctedAlpha); // 0.01 (0.05 / 5 tests)
 * console.log(result.significant);     // [true, false, false, false, false]
 * ```
 */
export declare const bonferroniCorrection: (pValues: number[], alpha?: number) => {
    correctedAlpha: number;
    significant: boolean[];
};
/**
 * Benjamini-Hochberg FDR correction.
 *
 * Controls false discovery rate (FDR) rather than FWER. Less conservative
 * than Bonferroni, making it more powerful for large numbers of tests.
 *
 * Procedure:
 * 1. Sort p-values in ascending order
 * 2. Find largest k such that p_k ≤ (k/m) * FDR
 * 3. Reject all hypotheses 1, 2, ..., k
 *
 * @param pValues - Array of p-values from multiple tests
 * @param fdr - Target false discovery rate (default: 0.05)
 * @returns Adjusted p-values and which tests are significant
 *
 * @example
 * ```typescript
 * const pValues = [0.001, 0.02, 0.045, 0.08, 0.12];
 * const result = benjaminiHochberg(pValues, 0.05);
 * console.log(result.adjustedPValues); // [0.005, 0.05, 0.075, 0.1, 0.12]
 * console.log(result.significant);     // [true, true, false, false, false]
 * ```
 */
export declare const benjaminiHochberg: (pValues: number[], fdr?: number) => {
    adjustedPValues: number[];
    significant: boolean[];
};
/**
 * Holm-Bonferroni method (step-down procedure).
 *
 * Less conservative than Bonferroni but still controls FWER.
 * More powerful than Bonferroni when not all hypotheses are rejected.
 *
 * Procedure:
 * 1. Sort p-values in ascending order
 * 2. Compare smallest p-value to α/m, next to α/(m-1), etc.
 * 3. Stop when you find first non-significant p-value
 *
 * @param pValues - Array of p-values from multiple tests
 * @param alpha - Original significance level (default: 0.05)
 * @returns Adjusted p-values and which tests are significant
 *
 * @example
 * ```typescript
 * const pValues = [0.001, 0.02, 0.045, 0.08, 0.12];
 * const result = holmBonferroni(pValues, 0.05);
 * console.log(result.significant); // [true, false, false, false, false]
 * ```
 */
export declare const holmBonferroni: (pValues: number[], alpha?: number) => {
    adjustedPValues: number[];
    significant: boolean[];
};
/**
 * Storey's q-value method (FDR-based with estimation of true nulls).
 *
 * More powerful than BH when many hypotheses are truly null.
 * Estimates π₀ (proportion of true null hypotheses) from data.
 *
 * @param pValues - Array of p-values from multiple tests
 * @param fdr - Target false discovery rate (default: 0.05)
 * @param lambda - Tuning parameter for π₀ estimation (default: 0.5)
 * @returns Q-values and which tests are significant
 *
 * @example
 * ```typescript
 * const pValues = [0.001, 0.02, 0.045, 0.08, 0.12, 0.35, 0.42, 0.55];
 * const result = storeyQValues(pValues, 0.05, 0.5);
 * console.log(result.qValues); // Estimated q-values
 * console.log(result.significant); // Which tests are significant at FDR=0.05
 * ```
 */
export declare const storeyQValues: (pValues: number[], fdr?: number, lambda?: number) => {
    qValues: number[];
    significant: boolean[];
    pi0: number;
};
//# sourceMappingURL=multiple-comparison.d.ts.map