[R-sig-ME] p-values in lme vs gamm

[Christoph Scherber]

Dear Etienne,

Given that the SE´s and the t values associated with the parameter estimates are identical in the lme and gamm.mod$gam fits, it seems that there are
differences in the calculation of the degrees of freedom. That is, the P values are calculated from the t distributions at different degrees of freedom.

Maybe some inspection of the GAMM source could might help.

All the best,
Christoph




Etienne Laliberté wrote:

> Thanks Christophe.
> 
> I'm aware it's useless to use gamm() when no smoothers are specified,
> but I still struggle to understand what exactly is the difference
> between summary(gamm.mod$gam) and summary(gamm.mod$lme).
> 
> Cheers
> 
> Etienne
> 
> Le lundi 05 juillet 2010 à 11:33 +0200, Christoph Scherber a écrit :
>> Dear Etienne,
>>
>> You should only compare the estimates and SE´s (and P values) from corresponding lme models:
>>
>> summary(lme.mod)
>> summary(gamm.mod$lme)
> 
>> These give exactly identical results.
>>
>> Looking at the gam model without explicitly specifying a smoothing function s() does not make too much sense.
>>
>> Best wishes
>> Christoph
>>
>>
>>
>>
>>
>> Etienne Laliberté wrote:
>>
>>> With gamm from the mgcv package, I'm struggling to understand the
>>> difference between summary(mymodel$gam) and summary(mymodel$lme).
>>>
>>> In particular, can someone enlighten me as to why some of the p-values
>>> differ with the following two calls? Without smoothers, I thought gamm
>>> would simply give the same results as lme().
>>>
>>> library(mgcv)
>>> data(oats)
>>> lme.mod <- lme(Y ~ N + V, random = list(B = ~1 , V = ~1), data = oats,
>>> method = "REML")
>>> gamm.mod <- gamm(Y ~ N + V, random = list(B = ~1 , V = ~1), data = oats,
>>> method = "REML")
>>> summary(lme.mod)
>>> summary(gamm.mod$gam)
>>> # note different p-values for VMarvellous and VVictory
>>>
>>> Thanks
>>>
>>> Etienne
>>>
>>> _______________________________________________
>>> R-sig-mixed-models at r-project.org mailing list
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>>>
>>> .
>>>
>>
> 
>