distrib_diff_lcwm {outlierMBC}R Documentation

Compute the dissimilarity for a linear cluster-weighted model and identify the lowest density observation.

Description

At each iteration of ombc_lcwm, distrib_diff_lcwm computes the dissimilarity value of the current linear cluster-weighted model. It also identifies the observation with the lowest mixture density.

Usage

distrib_diff_lcwm(x, z, prop, mu, sigma, mod_list, y_sigma, dd_weight = 0.5)

Arguments

x

Covariate data only.

z

Component assignment probability matrix.

prop

Vector of component proportions.

mu

Matrix of component mean vectors.

sigma

Array of component covariance matrices.

mod_list

List of component regression models.

y_sigma

Vector of component regression standard deviations.

dd_weight

A value between 0 and 1 which controls the weighting of the response and covariate dissimilarities when aggregating.

Value

distrib_diff_lcwm_lcwm returns a list with the following elements:

distrib_diff

Aggregated dissimilarity across components.

distrib_diff_vec

Vector containing dissimilarity value for each component.

choice_id

Index of observation with lowest mixture density.

removal_dens

Value of the lowest mixture density.

distrib_diff_mat

Two-column matrix containing response and covariate dissimilarities across components.


[Package outlierMBC version 0.0.1 Index]