[R-sig-ME] GLMM's with quadrature

Ben Bolker bbolker at gmail.com
Wed Dec 10 01:15:55 CET 2014


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 14-12-09 06:21 PM, Boylan, Ross wrote:
> I am interested in evaluating mixed models with * non-normal
> errors (binary and count outcomes) * random slopes and intercepts,
> possibly correlated (slopes and intercepts are over the same
> clusters) * quadrature
> 
> The documentation for lme4 indicates no restrictions on combining 
> these that I see, though I think in the past there were.  So, can
> I freely combine all 3 of the dimensions listed above?  Is there
> some other package worth considering?
> 
> There will also be some conventional LMM's.

  Yes, there is a restriction.  In previous (pre-1.0) versions of lme4
vector-valued random effects could be fitted by quadrature (but only
models with a single grouping variable could be used).  We still
haven't managed to get around to doing this for the modern version of
lme4: https://github.com/lme4/lme4/issues/123

  For LMMs, quadrature is not necessary; the integrals can be reduced
analytically to a single penalized, weighted residual sums of squares
computation that can be done without numerical integration or
approximations thereof.

> 
> Thanks. Ross Boylan
> 
> P.S. What's up with the FAQ?  The old one seems to have gone,
> though there is something identified as a draft at 
> http://glmm.wikidot.com/faq.

  Where was the old one?  That's the only one I'm aware of.

> _______________________________________________ 
> R-sig-mixed-models at r-project.org mailing list 
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
> 

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)

iQEcBAEBAgAGBQJUh5C7AAoJEOCV5YRblxUHkqQH/1mmz3GswBeQKFHTiqXP3cKZ
8IYjm6lJY3UGaH4Tu60BjbLIlg8gIjdkPater1yY92xfS9e6A6IOcVy2L7JWRwJ4
1PsdnISvovouFW1q/3nFg0Keo6FD+0JI7s2iSZ/u1pMf9KyHC9veQXpRAXolj3Zr
KsmRDr7v9bJIAJSO4NMQHsrDJ51png9boQmR7Gg+/x/EEXY+3DIAHDAxTsXG7rrp
DYkz0Tce8Gl5931HtliMXtB08v3Re23pZkNdlc8yLYTPj44xQ9MdNCnOtwMdEooV
lcrlcIrJwpTcK6pnE9lqxmipBoAEhoDETAHqtVu2Ea6rjiiFvkTv/0sg7jlm9nI=
=r5pO
-----END PGP SIGNATURE-----



More information about the R-sig-mixed-models mailing list