[R-sig-ME] Observation Level Random Effect in GLMM
davidD at qimr.edu.au
Tue May 17 02:58:49 CEST 2011
On Mon, 16 May 2011, Ross Culloch wrote:
> I was hoping someone could help me understand more about the use of the
> Observation-Level Random Effect.
> What I am not sure about is whether the OLRE takes in to account the
> variation in the explanatory variables, for example an individual may
> experience particularly extreme environmental conditions that other
> individuals did not experience due to temporal variation in their
> presence or absence at the study site, as a result there may be a
> greater (or lesser) variation within individual behaviour during such
> events. Is anyone aware of a reference that has used this approach in a
> paper and/or has explained what the OLRE is exactly doing - so that I
> can ease my mind on this approach?
Traditionally, it is seen as dealing with unmeasured explanatory
variables. Sometimes we have an indirect indicator of those variables eg
shared locality or parentage, sometimes we can only see an effect on the
overall distribution. Aitken et al (Statistical Modelling in GLIM) gives
a nice example by excluding an important measured covariate and showing
how this results in overdispersion. Breslow and Clayton's 1993 paper fits
several examples of different models, the Scottish lip cancer example is
useful (independent OLRE, spatially correlated OLREs, including the
possible environmental covariate).
Cheers, David Duffy.
| David Duffy (MBBS PhD) ,-_|\
| email: davidD at qimr.edu.au ph: INT+61+7+3362-0217 fax: -0101 / *
| Epidemiology Unit, Queensland Institute of Medical Research \_,-._/
| 300 Herston Rd, Brisbane, Queensland 4029, Australia GPG 4D0B994A v
More information about the R-sig-mixed-models