# Re: [R] Random seed problem in MCMC coupling of chains

From: Dimitris Rizopoulos <dimitris.rizopoulos_at_med.kuleuven.be>
Date: Wed 08 Jun 2005 - 23:53:37 EST

do you want something like this:

niter <- 3
nchain <- 2
rs <- sample(500, niter, TRUE)
for (i in 1:niter) { # iterations
for (j in 1:nchain) { # chains
set.seed(rs[i])
a <- runif(1)
cat("iter:", i, "chain:", j, "runif:", a, "\n")   }
}

I hope it helps.

Best,
Dimitris

Dimitris Rizopoulos
Ph.D. Student
Biostatistical Centre
School of Public Health
Catholic University of Leuven

Address: Kapucijnenvoer 35, Leuven, Belgium

```Tel: +32/16/336899
Fax: +32/16/337015
Web: http://www.med.kuleuven.ac.be/biostat/
http://www.student.kuleuven.ac.be/~m0390867/dimitris.htm

```
• Original Message ----- From: "Gorjanc Gregor" <Gregor.Gorjanc@bfro.uni-lj.si> To: <r-help@stat.math.ethz.ch> Sent: Wednesday, June 08, 2005 3:27 PM Subject: [R] Random seed problem in MCMC coupling of chains

> Hello!
>
> I am performing coupling of chains in MCMC and I need the same value
> of seed for two chains. I will show demo of what I want:
>
> R code, which might show my example is:
> niter <- 3
> nchain <- 2
> tmpSeed <- 123
> for (i in 1:niter) { # iterations
> for (j in 1:nchain) { # chains
> set.seed(tmpSeed)
> a <- runif(1)
> cat("iter:", i, "chain:", j, "runif:", a, "\n")
> tmpSeed <- .Random.seed
> }
> }
>
> I get this:
>
> iter: 1 chain: 1 runif: 0.43588
> iter: 1 chain: 2 runif: 0.43588
> iter: 2 chain: 1 runif: 0.43588
> iter: 2 chain: 2 runif: 0.43588
> iter: 3 chain: 1 runif: 0.43588
> iter: 3 chain: 2 runif: 0.43588
>
> but I would like to get:
>
> iter: 1 chain: 1 runif: 0.43588
> iter: 1 chain: 2 runif: 0.43588
> iter: 2 chain: 1 runif: 0.67676
> iter: 2 chain: 2 runif: 0.67676
> iter: 3 chain: 1 runif: 0.12368
> iter: 3 chain: 2 runif: 0.12368
>
> Note that seed value is of course changing, but it is parallel
> between chains.
>
> I am able to do only this, since setting seed at the beginning

> of chain i.e iteration is not a problem, but I want an upper
> scheme, since I compare chains and stop one if some condition is
> satisfied.
>
> tmpSeed <- 123
> for (i in 1:nchain) { # chains
> set.seed(tmpSeed)
> for (j in 1:niter) { # iterations
> a <- runif(1)
> cat("iter:", j, "chain:", i, "runif:", a, "\n")
> }
> }
> iter: 1 chain: 1 runif: 0.28758
> iter: 2 chain: 1 runif: 0.7883
> iter: 3 chain: 1 runif: 0.40898
> iter: 1 chain: 2 runif: 0.28758
> iter: 2 chain: 2 runif: 0.7883
> iter: 3 chain: 2 runif: 0.40898
> iter: 1 chain: 3 runif: 0.28758
> iter: 2 chain: 3 runif: 0.7883
> iter: 3 chain: 3 runif: 0.40898
>
> I was looking in 'rlecuyer', 'rsprng' and 'setRNG', but did not find
> anything usable for me. From reading on http://sprng.cs.fsu.edu/
> 'rsprng' provides just opposite of what I want, 'rlecuyer' is a bit
> to technical for me, but I think it also doesn't give identical
> seed for parallels. 'setRNG', especially it's function 'getRNG()'
> looks nice but its arguments should have seed stored. How can one
> do that?
>
>
> Thanks in advance!
>
> Lep pozdrav / With regards,
> Gregor Gorjanc
>
> ----------------------------------------------------------------------
> University of Ljubljana
> Biotechnical Faculty URI:
> http://www.bfro.uni-lj.si/MR/ggorjan
> Zootechnical Department mail: gregor.gorjanc <at> bfro.uni-lj.si
> Groblje 3 tel: +386 (0)1 72 17 861
> SI-1230 Domzale fax: +386 (0)1 72 17 888
> Slovenia, Europe
> ----------------------------------------------------------------------
> "One must learn by doing the thing; for though you think you know
> it,
> you have no certainty until you try." Sophocles ~ 450 B.C.
>
> ______________________________________________
> R-help@stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
> http://www.R-project.org/posting-guide.html
>

R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html Received on Thu Jun 09 00:00:34 2005

This archive was generated by hypermail 2.1.8 : Fri 03 Mar 2006 - 03:32:27 EST