Hi Ben, I suppose lme4 assumes the random effects has a multivariate normal distribution with mean vector 0 and a variance-covariance matrix. In that case, the marginal likelihood (obtained by integrating out the likelihood and random effect distributions) cannot be obtained in closed-form. I am not sure what type of approximations (quadrature, laplace, PQL etc.) do you use in lme4 for generalised linear mixed effects model? And do you have any good advice in choosing optimisers available in the package optimx? Thank you.
Approximation to Marginal Likelihood of GLMM in lme4
2 messages · Lee Yan Liang, Ken Beath
It is in the package documentation, either Laplace or adaptive Gauss-Hermite is available. For optimization methods use a quasi-Newton (BFGS) if it works, Nelder-Mead otherwise, I would avoid the others. These are questions you should have been able to answer yourself.
On 25 November 2014 at 02:24, Lee Yan Liang <yanlianglee at live.com> wrote:
Hi Ben,
I suppose lme4 assumes the random effects has a multivariate normal
distribution with mean vector 0 and
a variance-covariance matrix. In that case, the marginal likelihood
(obtained by integrating out the likelihood
and random effect distributions) cannot be obtained in closed-form.
I am not sure what type of approximations (quadrature, laplace, PQL etc.)
do you use in lme4 for generalised
linear mixed effects model? And do you have any good advice in choosing
optimisers available in the package
optimx?
Thank you.
[[alternative HTML version deleted]]
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
*Ken Beath* Lecturer Statistics Department MACQUARIE UNIVERSITY NSW 2109, Australia Phone: +61 (0)2 9850 8516 Building E4A, room 526 http://stat.mq.edu.au/our_staff/staff_-_alphabetical/staff/beath,_ken/ CRICOS Provider No 00002J This message is intended for the addressee named and may...{{dropped:9}}