robustbase: lmrob.fit – R documentation

Become an expert in R — Interactive courses, Cheat Sheets, certificates and more!

lmrob.fit

MM-type estimator for regression

Description

Compute MM-type estimators of regression: An S-estimator is used as starting value, and an M-estimator with fixed scale and redescending psi-function is used from there. Optionally a D-step (Design Adaptive Scale estimate) as well as a second M-step is calculated.

Usage

lmrob.fit(x, y, control, init = NULL, mf = NULL)

Arguments

`x`	design matrix (n x p) typically including a column of `1`s for the intercept.
`y`	numeric response vector (of length n).
`control`	a list of control parameters as returned by `lmrob.control`, used for both the initial S-estimate and the subsequent M- and D-estimates.
`init`	optional `list` of initial estimates. See Details.
`mf`	unused and deprecated.

Details

This function is the basic fitting function for MM-type estimation, called by lmrob and typically not to be used on its own.

If given, init must be a list of initial estimates containing at least the initial coefficients and scale as coefficients and scale. Otherwise it calls lmrob.S(..) and uses it as initial estimator.

Value

A list with components

`fitted.values`	X beta, i.e., `X %*% coefficients`.
`residuals`	the raw residuals, `y - fitted.values`
`rweights`	robustness weights derived from the final M-estimator residuals (even when not converged).
`rank`
`degree.freedom`	`n - rank`

`coefficients`	estimated regression coefficient vector
`scale`	the robustly estimated error standard deviation

`cov`	variance-covariance matrix of `coefficients`, if the RWLS iterations have converged (and `control$cov` is not `"none"`).
`control`

iter

`converged`	logical indicating if the RWLS iterations have converged.
`init.S`	the whole initial S-estimator result, including its own `converged` flag, see `lmrob.S` (only for MM-estimates).
`init`	A similar list that contains the results of intermediate estimates (not for MM-estimates).

Author(s)

Matias Salibian-Barrera, Martin Maechler and Manuel Koller