import { AnomalyPoint, AnomalyMethod } from './types';

/**
 * Z-Score method: Standard statistical approach
 * - Uses MEAN and STANDARD DEVIATION (parametric)
 * - Assumes normal distribution
 * - Sensitive to extreme outliers
 * - Best for: Clean, normally distributed data
 */
export declare function detectZScore(x: Float32Array | Float64Array, y: Float32Array | Float64Array, threshold?: number): AnomalyPoint[];
/**
 * MAD (Median Absolute Deviation) method: Robust statistical approach
 * - Uses MEDIAN instead of mean (non-parametric)
 * - Robust to outliers in the window
 * - Doesn't assume normal distribution
 * - Best for: Data with existing outliers, skewed distributions
 */
export declare function detectMAD(x: Float32Array | Float64Array, y: Float32Array | Float64Array, threshold?: number): AnomalyPoint[];
/**
 * IQR (Interquartile Range) method: Quartile-based approach
 * - Uses PERCENTILES (25th and 75th)
 * - Classic box-plot outlier detection
 * - Non-parametric, distribution-free
 * - Best for: General purpose, well-understood method
 */
export declare function detectIQR(x: Float32Array | Float64Array, y: Float32Array | Float64Array, multiplier?: number): AnomalyPoint[];
/**
 * Simplified Isolation Forest: Detects anomalies using random partitioning
 * - Machine learning approach
 * - Isolates anomalies through random splits
 * - Works well with high-dimensional data
 * - Best for: Complex patterns, unknown distributions
 */
export declare function detectIsolationForest(x: Float32Array | Float64Array, y: Float32Array | Float64Array, contamination?: number, numTrees?: number): AnomalyPoint[];
/**
 * Main detection function - routes to appropriate algorithm
 */
export declare function detectAnomalies(x: Float32Array | Float64Array, y: Float32Array | Float64Array, method: AnomalyMethod, sensitivity: number): AnomalyPoint[];