[R-sig-ME] Extracting the posterior distribution for a random effect in MCMCglmm
Jarrod Hadfield
j.hadfield at ed.ac.uk
Mon Aug 6 12:43:56 CEST 2012
Hi,
This:
>>> G-structure: ~day
>>> post.mean l-95% CI u-95% CI
>>> day 0.09326 0.06076 0.1313
summarises the posterior distribution of the variance of the normal
distribution from which the day *effects* come from. Samples from this
posterior distribution are stored in m2a.7$VCV[,"day"]. The effects
themselves are in Sol.
Cheers,
Jarrod
Quoting Robert Long <longrob604 at gmail.com> on Mon, 6 Aug 2012 11:28:04 +0100:
> Hi Jarrod, thanks for your reply.
>
> I understand about the interval calculation now. However, I'm sorry
> that I still don't see how to get the random effects myself. I see I
> can get the posterior mean for fixed effects by mean(m2a.7$Sol[,1])
> and HPDinterval(m2a.7$Sol[,1],prob=0.95) etc, I see there are data in
> m2a.7$Sol in columns after the fixed effects in columns 5 through 96
> for each of the days, but how do I reproduce this
>
>>> G-structure: ~day
>>> post.mean l-95% CI u-95% CI
>>> day 0.09326 0.06076 0.1313
>
> from m2a.7$Sol[,5:96]) ? I would like to do
> mean(something)
> and
> HPDinterval(something, prob = 0.95)
>
> So what is the "something" ?
>
> Thanks again
>
> Robert Long
> Postgraduate student
> University of Leeds / UK
>
> On Mon, Aug 6, 2012 at 11:11 AM, Jarrod Hadfield <j.hadfield at ed.ac.uk> wrote:
>> Hi,
>>
>> specifying pr=TRUE in the call to MCMCglmm saves the posterior distribution
>> of all location effects (fixed and random). They appear in Sol.
>>
>> summary uses HPDinterval not quantile. HPDinterval (with prob=0.95) finds
>> the shortest interval which contains 95% of the posterior samples, which may
>> be different from quantile which just finds the lowest and highest 2.5%.
>>
>> Cheers,
>>
>> Jarrod
>>
>>
>>
>>
>> Quoting Robert Long <longrob604 at gmail.com> on Mon, 6 Aug 2012 10:56:53
>> +0100:
>>
>>> Hello
>>>
>>> I would like to extract the data for the posterior distribution for a
>>> random effect in MCMCglmm. Using the example in the tutorial:
>>>
>>> data(Traffic)
>>> prior <- list(R = list(V = 1, nu = 0.002), G = list(G1 = list(V = 1,
>>> nu = 0.002)))
>>> m2a.7 <- MCMCglmm(y ~ year + limit + as.numeric(day), random = ~day,
>>> family = "poisson", data = Traffic, prior = prior, verbose = FALSE, pr=T)
>>>
>>> summary(m2a.7)
>>>
>>> This gives:
>>>
>>> G-structure: ~day
>>> post.mean l-95% CI u-95% CI eff.samp
>>> day 0.09326 0.06076 0.1313 266.8
>>>
>>> How can I extract the data that gives this mean and 95% BCI ?
>>>
>>> I can see that I can obtain the results for the fixed effects by such as:
>>> mean(m2a.7$Sol[,1]) which gives the posterior mean for the first fixed
>>> effect. But how can I do that for the random effects ? I can see that
>>> there are data in m2a.7$Sol[,5:96] but these don't seem to be
>>> variances as many are negative.
>>>
>>> A related question is: quantile(m2a.7$Sol[,1],c(0.025,0.975),type = 1)
>>> does not give precisely the same interval as in summary(m2a.7) - I
>>> wonder why there is a difference ?
>>>
>>> Thanks !
>>>
>>> Robert Long
>>> Postgraduate student
>>> University of Leeds / UK
>>>
>>> _______________________________________________
>>> 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.
>>
>>
>
>
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
More information about the R-sig-mixed-models
mailing list