convergence_precheck: Pre-check model matrices for convergence issues

In lmvar: Linear Regression with Non-Constant Variances

Description

The model matrices X_μ and X_σ are checked to see if problems with the convergence of the fit can be anticipated. If so, it is determined which columns must be removed from X_σ to attempt to avoid convergence issues.

Usage

convergence_precheck(y, X_mu, X_sigma)

Arguments

y	Numeric, response vector y
X_mu	Model matrix for the expected values
X_sigma	Model matrix for the standard deviations. This must be a full-rank matrix.

Details

A matrix can be of class 'matrix', 'Matrix' or 'numeric' (in case it is a matrix of one column only).

An intercept term must be included in the model matrices if the model is such.

Value

A list with the following members:

• column_numbers The numbers of the columns of X_sigma that can be kept

• column_names The names of the columns of X_sigma that can be kept

Numbers and names refer to the same columns. They are supplied both for convenience.

lmvar documentation built on May 16, 2019, 5:06 p.m.