[R-sig-ME] Questions on the results from glmmPQL(MASS)

[Ken Beath]

On 08/12/2008, at 6:42 PM, zhijie zhang wrote:

> On Mon, Dec 8, 2008 at 3:29 PM, David Duffy  
> <David.Duffy at qimr.edu.au> wrote:
>
>> On Mon, 8 Dec 2008, zhijie zhang wrote:
>>
>>
>>> Do u mean the following method?
>>>> model0<-glmmML(y ~ trt + I(week > 2), cluster=ID, family=binomial,
>>> data=bacteria)
>>>
>>>> model1<-glm(y ~ trt + I(week > 2), family=binomial, data=bacteria)
>>>> anova(model0,model1)
>>>>
>>> Error message occurred.
>>>
>>
>> anova does not have a method for glmmML, but the deviances seem to be
>> calculated the same (see model0$cluster.null.deviance etc):.
>>
>> model0 Residual deviance: 192.3  on 215 degrees of freedom       
>> AIC: 202.3
>> model1 Residual deviance: 199.18 on 216  degrees of freedom
>>
>>
>
>> LRTS = 6.88.  We will assume that the test statistic is distributed  
>> 1/2
>> X2(0) and 1/2 X2(1), so P ~ 0.004.
>
>
> 1/2 X2(0) and 1/2 X2(1): ?? what do they mean?
>

X2 is chi-square. Because the test is on the boundary the test  
statistic is distributed as the weighted sum of chi-square rather than  
the usual chi-square. Verbeke and Molenberghs cover this in their books.

Simulation (parametric bootstrap) seems a better way of doing this.  
Information criteria (AIC or BIC) can also be used. Most usual  
justification for a random effect is that there is expected to be one,  
so provided it can be estimated it is included.


>> Comparing to a Wald test using the SE on the random effect SD, I get:
>>
>> Z = 1.242/0.4024 = 3.08, P=0.001
>
> Is this a "clerical error"? Based on your hints, it seems that p  
> should be
> 0.003475077 .
>> dnorm(3.08)
> [1] 0.003475077
>   Thanks.
>

P value for z test is 2*dnorm(3.08) which is close enough to 0.001  
given that the test is only an approximation.

Ken


>>
>>
>>
>>
>>
>> --
>> | 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
>>
>
>
>
> -- 
> With Kind Regards,
>
> oooO:::::::::
> (..):::::::::
> :\.(:::Oooo::
> ::\_)::(..)::
> :::::::)./:::
> ::::::(_/::::
> :::::::::::::
> [***********************************************************************]
> ZhiJie Zhang ,PhD
> Dept.of Epidemiology, School of Public Health,Fudan University
> Office:Room 443, Building 8
> Office Tel./Fax.:+86-21-54237410
> Address:No. 138 Yi Xue Yuan Road,Shanghai,China
> Postcode:200032
> Email:epistat at gmail.com <Email%3Aepistat at gmail.com>
> Website: www.statABC.com
> [***********************************************************************]
> oooO:::::::::
> (..):::::::::
> :\.(:::Oooo::
> ::\_)::(..)::
> :::::::)./:::
> ::::::(_/::::
> :::::::::::::
>
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>