bugs-r-devel Mailing List for BUGS-R (Page 20)

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 422-6466

Hi,

I have looked at Dawn's change about "thinning" setup in R2WinBUGS and
would like to summarize it here, because it was not really clear (at 
least to me) what the change for S-PLUS side was all about.

Default behavior in R is now to create a BUGS script with the following
BUGS commands for a run with 5000 iterations, 1000 burn-in and thinning
interval 12 (showing only relevant parts):

---------------------------------------------------- BUGS script start

# after model, inits, ... set burnin phase

thin.updater (12)
update (84)

# set parameters to monitor

update(334)
coda()

------------------------------------------------------ BUGS script end

Dawn has changed this in S-PLUS part to:

---------------------------------------------------- BUGS script start

# after model, inits, ... set burnin phase

update (1000)
thin.updater (12)

# set parameters to monitor

update(4000)
coda()

------------------------------------------------------ BUGS script end

In my humble opinion this is better:
  - BUGS script is clearer
  - CODA files have indices that match iteration number

This does not affect R2WinBUGS as it has only effect on produced CODA
files. For example, reading CODA files into R from above runs, would
after read.coda() give the following info:

# Current implementation
Iterations = 85:418
Thinning interval = 1
Number of chains = 1
Sample size per chain = 334

# Proposed implementation by Dawn
Iterations = 1001:4985
Thinning interval = 12
Number of chains = 1
Sample size per chain = 333

Better, again!

If there are no objections, I will commit this change.

-------- Original Message --------
Date: Thu, 19 Apr 2007 18:39:50 -0400
From: Andrew Gelman <ge...@st...>

G
I don't actually understand the question!
A

Gregor Gorjanc wrote:
> Hi,
>
> In my opinion Dawn's change make sense also for R side. Does anyone have
> any objection for not porting this also to R side?
>
> Gregor
>
> -------- Original Message --------
> Subject: [Bugs-r-devel] SF.net SVN: bugs-r: [12]
> trunk/R2WinBUGS/R/bugs.script.R
> Date: Tue, 17 Apr 2007 07:21:52 -0700
> From: woo...@us...
> To: bug...@li...
>
> Revision: 12
>           http://svn.sourceforge.net/bugs-r/?rev=12&view=rev
> Author:   woodard_
> Date:     2007-04-17 07:21:52 -0700 (Tue, 17 Apr 2007)
>
> Log Message:
> -----------
> In S-PLUS, the handling of the thinning is done differently than in R.
> bin represents the number of iterations between saves, before thinning,
> where in R it is the number of iterations between saves, after thinning.
> This alternative handling of the thinning is done so that the resulting
> samples have the correct iteration indexes in the output (coda) files.
> Therefore, if the samples are read into S-PLUS using the coda package,
> the thinning will be correctly labelled in the resulting mcmc object.
> In R, the thinning is always labelled as 1, even if thinning was done.
>
> Modified Paths:
> --------------
>     trunk/R2WinBUGS/R/bugs.script.R
>
> Modified: trunk/R2WinBUGS/R/bugs.script.R
> ===================================================================
> --- trunk/R2WinBUGS/R/bugs.script.R	2007-04-16 11:21:38 UTC (rev 11)
> +++ trunk/R2WinBUGS/R/bugs.script.R	2007-04-17 14:21:52 UTC (rev 12)
> @@ -21,6 +21,24 @@
>    initlist <- paste("inits (", 1:n.chains, ", '", inits, "')\n", sep="")
>    savelist <- paste("set (", parameters.to.save, ")\n", sep="")
>    redo <- ceiling((n.iter-n.burnin)/(n.thin*bin))
> +
> +  if (is.R()){
> +    thinUpdateCommand <- paste("thin.updater (", n.thin, ")\n",
> +      "update (", ceiling(n.burnin/n.thin), ")\n", sep = "")
> +  } else{
> ...
> +  	thinUpdateCommand <- paste("update (", n.burnin, ")\n",
> +			"thin.samples (", n.thin, ")\n", sep = "")
> +		bin = bin * n.thin
> +  }
> +
>    cat(
>      "display ('log')\n",
>      "check ('", native2win(model), "')\n",
> @@ -28,8 +46,7 @@
>      "compile (", n.chains, ")\n",
>      if(is.inits) initlist,
>      "gen.inits()\n",
> -    "thin.updater (", n.thin, ")\n",
> -    "update (", ceiling(n.burnin/n.thin), ")\n",
> +    thinUpdateCommand,
>       savelist,
>      if(DIC) "dic.set()\n",
>      rep(
>
>
> This was sent by the SourceForge.net collaborative development platform,
> the world's largest Open Source development site.
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by DB2 Express
> Download DB2 Express C - the FREE version of DB2 express and take
> control of your XML. No limits. Just data. Click to get it now.
> http://sourceforge.net/powerbar/db2/
> _______________________________________________
> Bugs-r-devel mailing list
> Bug...@li...
> https://lists.sourceforge.net/lists/listinfo/bugs-r-devel
>
>   

-- 
Andrew Gelman
Professor, Department of Statistics
Professor, Department of Political Science
Columbia University, New York
ge...@st...
www.stat.columbia.edu/~gelman

Office hours spring 2006:
  to be announced

Statistics department office:
  Social Work Bldg (Amsterdam Ave at 122 St), Room 1016
  212-851-2142
Political Science department office:
  International Affairs Bldg (Amsterdam Ave at 118 St), Room 731
  212-854-7075

Mailing address:
  1255 Amsterdam Ave, Room 1016
  Columbia University
  New York, NY 10027-5904
  212-851-2142
  (fax) 212-851-2164

-- 
Lep pozdrav / With regards,
     Gregor Gorjanc
----------------------------------------------------------------------
University of Ljubljana     PhD student
Biotechnical Faculty        www: http://www.bfro.uni-lj.si/MR/ggorjan
Zootechnical Department     blog: http://ggorjan.blogspot.com
Groblje 3                   mail: gregor.gorjanc <at> bfro.uni-lj.si
SI-1230 Domzale             fax: +386 (0)1 72 17 888
Slovenia, Europe            tel: +386 (0)1 72 17 861
----------------------------------------------------------------------

2006	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov (11)	Dec (20)
2007	Jan	Feb	Mar	Apr (7)	May (24)	Jun (15)	Jul (5)	Aug (21)	Sep (85)	Oct (9)	Nov	Dec
2008	Jan (15)	Feb (11)	Mar	Apr (2)	May (3)	Jun	Jul	Aug	Sep (29)	Oct (47)	Nov (1)	Dec (1)
2009	Jan (8)	Feb (12)	Mar (14)	Apr (4)	May (1)	Jun	Jul (5)	Aug	Sep	Oct (11)	Nov (18)	Dec
2010	Jan	Feb	Mar (1)	Apr (3)	May (24)	Jun	Jul (1)	Aug (7)	Sep (5)	Oct (3)	Nov (4)	Dec
2011	Jan	Feb	Mar (13)	Apr (3)	May (2)	Jun	Jul (1)	Aug (24)	Sep (3)	Oct	Nov	Dec (29)
2012	Jan (5)	Feb (5)	Mar	Apr	May (1)	Jun	Jul (3)	Aug	Sep	Oct	Nov	Dec
2013	Jan	Feb	Mar	Apr (9)	May (1)	Jun	Jul	Aug (2)	Sep	Oct	Nov	Dec
2015	Jan	Feb	Mar	Apr (1)	May (1)	Jun	Jul (3)	Aug	Sep	Oct	Nov	Dec (1)
2017	Jan	Feb	Mar	Apr	May	Jun (5)	Jul	Aug	Sep	Oct	Nov	Dec