Re: [Bugs-r-devel] as.mcmc.list

[Mike M. <mme...@wc...>]

Hi Uwe and Gorjanc,

Looks like I got the thinning wrong!

Thanks for correcting that.

Cheers, Mike.

At 19:52 20/02/2009, Uwe Ligges wrote:
>Gorjanc Gregor wrote:
>>Hi,
>>Recently as.mcmc.list.bugs has been introduced as contrbuted by 
>>Mike Meredith.
>>I think there are some flaws. For example. I have an bugs object 
>>with the following details:
>>3 chains, each with 1e+05 iterations (first 10000 discarded), 
>>n.thin = 10 n.sims = 27000 iterations saved
>>i.e.
>>List of 24
>>  $ n.chains       : num 3
>>  $ n.iter         : num 1e+05
>>  $ n.burnin       : num 10000
>>  $ n.thin         : num 10
>>  $ n.keep         : num 9000
>>  $ n.sims         : num 27000
>>however, when I create an mcmc object I get
>>List of 3
>>  $ : mcmc [1:9000, 1:59] 12 12.4 12.8 13.1 12.1 ...
>>   ..- attr(*, "dimnames")=List of 2
>>   .. ..$ : chr [1:9000] "1001" "1002" "1003" "1004" ...
>>   .. ..$ : chr [1:59] "alpha" "deviance" "l[10]" "l[11]" ...
>>   ..- attr(*, "mcpar")= num [1:3] 1001 10000 1
>>...
>>The last line, i.e., the mcpar, is wrong. It should be
>>   ..- attr(*, "mcpar")= num [1:3] 10001 100000 10
>>I also do not see the point in setting the rownames, since they are 
>>not reflecting the reality, i.e.,
>>my chains where thinned, but the rownames are 1001, 1002, ...
>>Bellow is the diff of the "fixed" function, but before I commit it, 
>>I would like to make sure
>>there is consensus about this. I could not find Mikes e-mail. Uwe?
>
>Thanks for the fix. I had not tested all possible cases, obviously. 
>I'm fine with it, hence just commit unless anybody else objects. CCing Mike.
>
>Best,
>Uwe
>
>
>
>>Index: as.mcmc.list.bugs.R
>>===================================================================
>>--- as.mcmc.list.bugs.R (revision 105)
>>+++ as.mcmc.list.bugs.R (working copy)
>>@@ -5,14 +5,14 @@
>>      if(dim(x$sims.array)[2] != x$n.chains)
>>          stop("Inconsistancy in bug object regarding the number of chains.")
>>      mclis <- vector("list", x$n.chains)
>>-    strt <- x$n.burnin%/%x$n.thin + 1
>>-    end <- x$n.iter%/%x$n.thin
>>+    strt <- x$n.burnin + 1
>>+    end <- x$n.iter
>>      ord <- order(dimnames(x$sims.array)[[3]])
>>-    if(end - strt + 1 < nrow(x$sims.array[,1,ord])) end <- end + 1
>>+    ## This is not necedarry, since strt and end match dims of sims.array
>>+    ## if(end - strt + 1 < nrow(x$sims.array[,1,ord])) end <- end + 1
>>      for(i in 1:x$n.chains) {
>>          tmp1 <- x$sims.array[,i,ord]
>>-        rownames(tmp1) <- strt:end
>>-        mclis[[i]] <- mcmc(tmp1, strt, end, 1)
>>+        mclis[[i]] <- mcmc(tmp1, start=strt, end=end, thin=x$n.thin)
>>------------------------------------------------------------------------------
>>Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
>>-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
>>-Strategies to boost innovation and cut costs with open source participation
>>-Receive a $600 discount off the registration fee with the source code: SFAD
>>http://p.sf.net/sfu/XcvMzF8H
>>_______________________________________________
>>Bugs-r-devel mailing list
>>Bug...@li...
>>https://lists.sourceforge.net/lists/listinfo/bugs-r-devel
>
>===============================================
>Michael E Meredith
>Wildlife Conservation Society (WCS) Malaysia Program
>7 Jalan Ridgeway, 93250 Kuching, Malaysia
>Fax: +60-82-577 443 Mobile: +60-19 888 1533
>email: mme...@wc...
>Program website: http://www.wcsmalaysia.org