Сравнение методов

	Оценка на основе медианного абсолютного отклонения (MAD)	Квантильная регрессия	Гребневая регрессия
Область	Статистика	Эконометрика	Машинное обучение
Семейство	Regression model	Regression model	Machine learning
Год появления	1974	1978	1970
Автор метода	Hampel (influence-curve treatment); classical robust statistics	Koenker & Bassett	Hoerl, A.E. & Kennard, R.W.
Тип	Robust scale estimator	Conditional quantile regression	L2-regularized linear regression
Основополагающий источник	Hampel, F. R. (1974). The Influence Curve and Its Role in Robust Estimation. Journal of the American Statistical Association, 69(346), 383-393.	Koenker, R. & Bassett, G., Jr. (1978). Regression Quantiles. Econometrica, 46(1), 33-50.	Hoerl, A.E. & Kennard, R.W. (1970). Ridge Regression: Biased Estimation for Nonorthogonal Problems. Technometrics, 12(1), 55–67.
Другие названия	median absolute deviation, MAD scale estimator, robust scale estimation, Medyan Mutlak Sapma (MAD) Tahmini	conditional quantile regression, regression quantiles, Kantil Regresyon	Ridge Regresyonu, ridge regresyonu, L2-regularized regression, Tikhonov regularization
Связанные	5	5	4
Сводка	Median Absolute Deviation estimation is a robust measure of statistical dispersion that replaces the standard deviation when outliers are present. Rooted in the influence-curve framework formalised by Hampel (1974), it summarises the spread of a continuous variable using medians instead of means, so a single extreme value cannot distort the result.	Quantile regression models conditional quantiles of an outcome - the median, the 25th or 75th percentile, and so on - rather than the conditional mean that OLS targets. Introduced by Koenker and Bassett in 1978, it reveals how predictors act across the whole distribution, including its tails.	Ridge Regression is an L2-regularized linear regression method, introduced by Arthur Hoerl and Robert Kennard in 1970, that reduces multicollinearity by adding a penalty on the size of the coefficients. It shrinks coefficients toward zero without setting any of them exactly to zero, producing more stable estimates when predictors are highly correlated.