simple_EM

EM_prelim_logit

EM_prelim_nb

Use ridge penalty to prevent separation. Not be called by user!

internal

Estimates hierarchical models using mean-field variational Bayes.
At present, it can estimate logistic, linear, and negative binomial models.
It can accommodate models with an arbitrary number of random effects and
requires no integration to estimate. It also provides the ability to improve
the quality of the approximation using marginal augmentation.
Goplerud (2022) <doi:10.1214/21-BA1266> and Goplerud (2024) <doi:10.1017/S0003055423000035>
provide details on the variational algorithms.

Max Goplerud

vglmer

Variational Inference for Hierarchical Generalized Linear Models

simple_EM function

<dl><dt>X</dt>
<dd>Design matrix</dd>
<dt>Z</dt>
<dd>RE design matrix</dd>
<dt>s</dt>
<dd>(y_i - n_i)/2 for polya-gamma input</dd>
<dt>pg_b</dt>
<dd>n_i as vector input</dd>
<dt>iter</dt>
<dd>iterations</dd>
<dt>ridge</dt>
<dd>variance of ridge prior</dd>
<dt>y</dt>
<dd>Raw observed y_i</dd>
<dt>est_r</dt>
<dd>Initial r value (not updated!)</dd></dl>

Arguments

Simple EM algorithm for starting values. — simple_EM

<dl>

<dt>X</dt>
<dd>Design matrix</dd>


<dt>Z</dt>
<dd>RE design matrix</dd>


<dt>s</dt>
<dd>(y_i - n_i)/2 for polya-gamma input</dd>


<dt>pg_b</dt>
<dd>n_i as vector input</dd>


<dt>iter</dt>
<dd>iterations</dd>


<dt>ridge</dt>
<dd>variance of ridge prior</dd>


<dt>y</dt>
<dd>Raw observed y_i</dd>


<dt>est_r</dt>
<dd>Initial r value (not updated!)</dd>

</dl>

simple_EM: Simple EM algorithm for starting values.

Description

Usage

Arguments