[R-sig-ME] Fitting known var-cov matrix in logistic regression model

[Jarrod Hadfield]

Hi Margaret,

I'm in the middle of field work so the answer is short - sorry.

Vsvd<-svd(VCV)  # VCV is your var-cov matrix for populations
Vsvd<-Vsvd$v%*%(t(Vsvd$u)*sqrt(Vsvd$d))
Z<-model.matrix(~pop-1, gpdata)%*%Vsvd

gpdata$Z<-Z

out<-MCMCglmm(cbind(pos,neg)~1+env,random=~idv(Z)pop,data=gpdata,family="multinomial2",prior=prior,burnin=3000,nitt=20000)

For zibinomial read the section on zipoisson in the CourseNotes (or  
look at previous posts) for zipossion. The latent variables are set up  
in the same way.

Jarrod







Quoting Dominick Samperi <djsamperi at gmail.com> on Sat, 23 Apr 2011  
17:53:52 -0400:

> On Fri, Apr 22, 2011 at 2:31 PM, Margaret Mackinnon
> <mmackinnon at kilifi.kemri-wellcome.org> wrote:
>> Dear David
>>
>> Thanks for your response.
>>
>> I tried geeglm and glmPQL and MCMCglmm.  I still can't work out how  
>> to fit the fixed var-cov structure. Here is what I did for MCMCglmm  
>> (my preferred option):
>>
>>  gpdata<-data.frame(pos=as.integer(freq*n*2),neg=as.integer(n*2-freq*n*2),pop=c(1:15),env)
>> prior<-list(R=list(V=diag(1),nu=0.002),G=list(G1=list(V=10,nu=0.002)))
>> out<-MCMCglmm(cbind(pos,neg)~1+env,random=~pop,data=gpdata,family="multinomial2",prior=prior,burnin=3000,nitt=20000)
>> summary(out$VCV)
>> plot(out$VCV)
>> plot(out$Sol)
>> summary(out$Sol)
>>
>> This gives sensible answers for the random effect (pop) variance  
>> and the fixed effects (intercept + env), i.e. those given by lmer  
>> and glm.  However I can't understand the manual on how to fit a  
>> predfined value of the var-cov matrix that describes the different  
>> variances for each pop and the covariances between them.  Is it in  
>> rcov, or in the priors (R or G?) with the fix option on.
>
> This is an interesting question. I think packages like nlme permit you
> to define the
> form of the var-cov matrix, but not to specify it completely. Thus
> there is a trade-off
> between how much domain knowledge you can inject and how much the software
> will discover in the data. If I understand the docs correctly any
> specified values for
> the var-cov params are used as initial conditions, not as fixed values.
>
> Dominick
>
>> I also could not get the family="zibinomial" to work because it  
>> needs idh or us variance functions which I don't know how to set up  
>> for this particular problem.
>>
>> I would really appreciate some help on this.
>>
>> Margaret
>>
>>
>>
>>
>>
>>
>>>>> "David Duffy" <davidD at qimr.edu.au> 22 April 2011 04:49 >>>
>> On Thu, 21 Apr 2011, Margaret Mackinnon wrote:
>>
>>> I want to estimate the relationship between population allele
>>> frequencies at a certain locus of interest and an environmental
>>> variable.  The hypothesis is that this environmental variable has
>>> generated these different allele frequencies through differential
>>> selection pressure on a locus under balancing selection (i.e., there is
>>> negative selection which balances the differential positive selection on
>>> the allele of interest).   The allele frequencies are measured in 15
>>> different populations, as are the environmental variables.  However,
>>> this environmental variable has been aggregated at the population level
>>> so there are only 15 values that it can take.
>>>
>>>  The populations are genetically related to each other to different
>>> degrees and so I want to reflect this structuring by including in the
>>> model a variance-covariance matrix which captures this relatedness.  I
>>> estimated this variance-covariance matrix from a set of independent data
>>> on SNPs at a whole lot of putatively neutral loci that are not expected
>>
>> The easiest way might be to use the SNP principal component scores for the
>> populations as covariates in an ordinary logistic regression. Otherwise
>> a GEE with your specified matrix, glmmPQL with a corSymm as per your model,
>> MCMCglmm.
>>
>> --
>> | 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
>>
>> ______________________________________________________________________
>> This email has been scanned by the MessageLabs Email Security System.
>> For more information please visit http://www.messagelabs.com/email
>> ______________________________________________________________________
>>
>> This e-mail (including any attachment to it) contains information  
>> which is confidential. It is intended only for the use of the named  
>> recipient. If you have received this e-mail in error, please let us  
>> know by replying to the sender, and immediately delete it from your  
>> system.  Please note, that in these circumstances, the use,  
>> disclosure, distribution or copying of this information is strictly  
>> prohibited. We apologize for any inconvenience that may have been  
>> caused to you. KEMRI-Wellcome Trust Programme cannot accept any  
>> responsibility for the  accuracy or completeness of this message as  
>> it has been transmitted over a public network. KEMRI-Wellcome Trust  
>> Programme reserves the right to monitor all incoming and outgoing  
>> email traffic. Although the Programme has taken reasonable  
>> precautions to ensure no viruses are present in emails, it cannot  
>> accept responsibility for any loss or damage arising from the use  
>> of the email or attachments. Any views expressed in this message are!
>>  those of the individual sender, except where the sender  
>> specifically states them to be the views of KEMRI-Wellcome Trust  
>> Programme.
>>
>> _______________________________________________
>> R-sig-mixed-models at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>>
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>
>



-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.