Tonker wrote:>
> Hi
> I am trying to run an lme on count data.
>
> lme (fixed = lognobees ~ round + spray + dist + fenconcpoint +
> uncolt200mbuffer + daysafterspray + round:spray + round:fenconcpoint +
> spray:fenconcpoint + spray:uncolt200mbuffer + dist:uncolt200mbuffer +
> uncolt200mbuffer:daysafterspray, random=~1|field/transect/subj,
> method="ML")
>
> The model runs ok but my data is count data and i need to model it with a
> poisson - in this case quasipoisson as its overdispersed distribution or a
> negative binomial distirubtion. While i know this is possible in lmer with
> family=quasipoisson i cannot find how to specify this in lme and my model
> doesnt work in lmer.
>
> Please could someone let me know if it is possible and if so how to state
> the family distribution in lme.
>
>
Not possible. You could try glmmPQL [MASS package] or the glmmADMB
package ... or, as a hack, use the weights argument to allow for
heteroscedasticity.
Ben Bolker
--
View this message in context:
http://www.nabble.com/defining-error-families-in-lme-%21-tf4835069.html#a13835832
Sent from the R help mailing list archive at Nabble.com.