Skip to content

user-defined covariance matrix for linear mixed models

5 messages · philippe fullsack, Vincent Careau, David Duffy

Original
Vincent Careau
# 
Hi,
I would say ASReml-R. It is free for academics using Windows. I don't know
if someone else figured out how to do an animal model (that seems like what
you want to do - or something very similar) with other mixed model packages
in R.  ASReml-R uses the Average Information algorithm. Perhaps the typical
methods used for the R packages would take too long with as many random
effects as an animal model is trying to estimate. 

By the way, Matt Wolak just created the "nadiv" package that you can use to
include dominance in your animal model, given you have the pedigree links to
estimate it independently from additive genetic variance. 
Good luck,
vc

Vincent Careau
NSERC Postdoctoral fellow
Department of Biology
University of California
Riverside, CA, 92521
http://faculty.ucr.edu/~vcareau/


-----Original Message-----
From: r-sig-mixed-models-bounces at r-project.org
[mailto:r-sig-mixed-models-bounces at r-project.org] On Behalf Of philippe
fullsack
Sent: Monday, October 01, 2012 5:58 AM
To: r-sig-mixed-models at r-project.org
Subject: [R-sig-ME] user-defined covariance matrix for linear mixed models

I am looking for an R package or function that would estimate the
variance-covariance matrix of a linear mixed model.
I have used pedigreemm and lme4 for simple examples but I would like to
know how to provide my own covariance structure.
E.g. G, the variance-covariance matrix of a lmm model will typically be the
tensor product of a typically small matrix g (e.g. a 2 x2 matrix of 3
variance-covariance parameters to estimate : sigxx,sigxy,sigyy) with of a
known matrix A. I wish to read matrix A from a file, and ask the R function
to estimate g. I am not interested only in the case of diagonal g matrices.

Is anybody aware of an interface e.g. to lme4, that would allow users to
specify and fit their models in such, flexible, way?
Please specify in your answer the input format for A (i.e sparse CSR,
dense, trios, or any) - (g is in dense format).

Note that a package like Jarrod Hadfield's MCMCglmm allows such flexible
definitions. However, I am looking for a non MCMC estimator (e.g. REML).

P.Fullsack
Dalhousie University


_______________________________________________
R-sig-mixed-models at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
David Duffy
#
On Mon, 1 Oct 2012, philippe fullsack wrote:

            

      
      
      
    

        
library(help=regress)


Version:              1.3-8
Date:                 2012-03-19
Title:                Gaussian linear models with linear covariance
                       structure
Author:               David Clifford and Peter McCullagh. Additional
                       contributions by HJ Auinger.
Maintainer:           David Clifford <david.clifford at csiro.au>
Description:          Functions to fit Gaussian linear model by
                       maximising the residual log likelihood where the
                       covariance structure can be written as a linear
                       combination of known matrices.  Can be used for
                       multivariate models and random effects models.
                       Easy straight forward manner to specify random
                       effects models, including random interactions.