[R-sig-ME] problem using lmer with family = Gamma

[Rafael Mares]

Hi David

Thank you very much for your reply.  I have tried the options you
suggest, and it seems to me that the Box-Cox transformation does a
good job of normalizing the data, but I would rather not transform the
response variable, based on Zuur et al's data exploration protocol
(2010) - which seems to suggest that in my case, differences in time
spent sniffing between the control and treatment may be made smaller
after transformation.

Do you think a transformation would be preferable over using Gamma
errors in my case?  Possibly a difficult question to answer not having
seen the raw data, but maybe there is something about mixed models in
lmer with Gamma errors that I am completely missing.

Thanks again for your help.
All the best,


Raff



On 12 June 2010 00:55, David Duffy <davidD at qimr.edu.au> wrote:
> On Fri, 11 Jun 2010, Rafael Mares wrote:
>
>> Dear all
>>
>> I'm trying to run a GLMM with Gamma errors (as seems most appropriate
>> for my data) using the lme4 package (version 0.999375-33)
>>
>> fullmod<-lmer(sniff ~ treatment + sex + domstatus + donor.age +
>> (1|group/id), family = Gamma, REML = FALSE, data = pres)
>>
>> Error in mer_finalize(ans) : mu[i] must be positive: mu = -44.2351, i = 1
>
> May I ask if you have fitted an ordinary GLM using the fixed effects? Have
> you perhaps tried a Box-Cox approach, or a LMM with log(sniff)?
>
> 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
>



-- 
Rafael Mares
Large Animal Research Group (LARG)
Department of Zoology
University of Cambridge
Downing Street
Cambridge
CB2 3EJ