GLVmix: NPMLE of Gaussian Location-Scale Mixture Model

In REBayes: Empirical Bayes Estimation and Inference

NPMLE of Gaussian Location-Scale Mixture Model

Description

A Kiefer-Wolfowitz procedure for ML estimation of a Gaussian model with possibly dependent mean and variance components. This version differs from WGLVmix in that it doesn't assume the data is in longitudinal form. This version assumes a general bivariate distribution for the mixing distribution. The defaults use a rather coarse bivariate gridding.

Usage

GLVmix(t, s, m, u = 30, v = 30, ...)

Arguments

t	A vector of location estimates
s	A vector of variance estimates
m	A vector of sample sizes of the same length as t and s, or if scalar a common sample size length
u	A vector of bin boundaries for the location effects
v	A vector of bin boundaries for the variance effects
...	optional parameters to be passed to KWDual to control optimization

Value

A list consisting of the following components:

u	midpoints of mean bin boundaries
v	midpoints of variance bin boundaries
fuv	the function values of the mixing density.
logLik	log likelihood value for mean problem
du	Bayes rule estimate of the mixing density means.
dv	Bayes rule estimate of the mixing density variances.
A	Constraint matrix
status	Mosek convergence status

Author(s)

R. Koenker and J. Gu

References

Gu, J. and R. Koenker (2014) Heterogeneous Income Dynamics: An Empirical Bayes Perspective, JBES,35, 1-16.

Koenker, R. and J. Gu, (2017) REBayes: An R Package for Empirical Bayes Mixture Methods, Journal of Statistical Software, 82, 1–26.

See Also

WTLVmix for an implementation assuming independent heterogeneity, and WGLVmix for a version that requires access to a full longitudinal data structure.

REBayes documentation built on Aug. 19, 2023, 5:10 p.m.