Skip to content

GLMMs with Adaptive Gaussian Quadrature - GLMMadaptive 0.4-0

1 message · Dimitris Rizopoulos

Original
Dimitris Rizopoulos
# 
Dear R mixed-model users,

A new version of GLMMadaptive (0.4-0) has been rolled out on CRAN.

Summary: GLMMadaptive can fit mixed effects models using adaptive 
Gaussian quadrature to approximate the integrals over the random 
effects, allowing also for user-specified models.

New features:

- New family objects and extended support for models with an extra 
zero-part:

   * Zero-inflated Poisson using zi.poisson(), and hurdle/truncated 
Poisson using hurdle.poisson().

   * Zero-inflated negative binomial using zi.negative.binomial(), and 
hurdle/truncated negative binomial using hurdle.negative.binomial().

   * Two-part/hurdle model for semi-continuous Gaussian outcomes using 
hurdle.lognormal().

   * Two-part/hurdle Beta using hurdle.beta.fam().

   * More info in: vignette("ZeroInflated_and_TwoPart_Models", package = 
"GLMMadaptive")

- Full support has been added for multiple comparisons and effects 
estimates as provided the **emmeans** and **multcomp** packages. More 
info in: vignette("Multiple_Comparisons", package = "GLMMadaptive")

- The predict() method

   * now works also for models with an extra zero-part;

   * it has been extended to calculate dynamic predictions for the 
different types of models using the 'newdata2' argument;

   * the function scoring_rules() calculate proper scoring rules for 
categorical data.

   * More info in: vignette("Dynamic_Predictions", package = "GLMMadaptive")

As always, any kind of feedback is more than welcome.

Best,
Dimitris