R2WinBUGS error in R

Question

I'm trying to duplicate some code and am running into troubles with WinBUGS. The code was written in 2010 and I think that back then, the package was installed with additional files which R is now looking for and can't find (hence the error), but I'm not sure.

R stops trying to run #bugs.directory (see code) and the error is:

Error in file(con, "rb") : cannot open the connection
In addition: Warning message:
In file(con, "rb") :
  cannot open file 'C:/Users/Hiwi/Documents/R/Win-library/3.0/R2winBUGS/System/Rsrc/Registry.odc': No such file or directory
Error in bugs.run(n.burnin, bugs.directory, WINE = WINE, useWINE = useWINE,  : 
  WinBUGS executable does not exist in C:/Users/Hiwi/Documents/R/Win-library/3.0/R2winBUGS

I have the results of the analysis so if there is another way of conducting a Bayesian analysis for the "rawdata" file (in the 14 day model with [-3,0] event window) or if someone would PLEASE shed some light on what's wrong with the code, I would be forever grateful.

The code is:

rm(list=ls(all=TRUE))   

setwd("C:/Users/Hiwi/Dropbox/Oracle/Oracle CD files/analysis/chapter6_a")

library(foreign)

rawdata <- read.dta("nyt.dta",convert.factors = F)

library(MASS)
summary(glm.nb(rawdata$num_events_14 ~ rawdata$nyt_num))


# WinBUGS code

library("R2WinBUGS")
nb.model <- function(){
        for (i in 1:n){ # loop for all observations 
               # stochastic component
               dv[i]~dnegbin( p[i], r)
               # link and linear predictor
               p[i] <- r/(r+lambda[i])
               log(lambda[i] ) <- b[1] + b[2] * iv[i]
        }
        #
        # prior distributions
        r <- exp(logr)
    logr ~ dnorm(0.0, 0.01)

    b[1]~dnorm(0,0.001)  # prior  (please note: second element is 1/variance)
    b[2]~dnorm(0,0.001)  # prior
}


write.model(nb.model, "negativebinomial.bug")


n <- dim(rawdata)[1] # number of observations


winbug.data <- list(dv = rawdata$num_events_14, 
                    iv = rawdata$nyt_num,
                    n=n)

winbug.inits <- function(){list(logr = 0 ,b=c(2.46,-.37)
                                )} # Ausgangswerte aus der Uniformverteilung zwischen -1 und 1

bug.erg <- bugs(data=winbug.data,
    inits=winbug.inits, 
    #inits=NULL,
    parameters.to.save = c("b","r"),
    model.file="negativebinomial.bug",
    n.chains=3, n.iter=10000, n.burnin=5000,
    n.thin=1,
    codaPkg=T,
    debug=F,
    #bugs.directory="C:/Users/Hiwi/Documents/R/Win-library/3.0/R2winBUGS/"
    bugs.directory="C:/Users/Hiwi/Documents/R/Win-library/3.0/R2winBUGS"
    )


tempdir()
setwd(tempdir())
file.rename("codaIndex.txt","simIndex.txt")
file.rename("coda1.txt","sim1.txt")
file.rename("coda2.txt","sim2.txt")
file.rename("coda3.txt","sim3.txt")

posterior <- rbind(read.coda("sim1.txt","simIndex.txt"),read.coda("sim2.txt","simIndex.txt"),read.coda("sim3.txt","simIndex.txt"))
post.df <- as.data.frame(posterior)
summary(post.df)
quantile(post.df[,2],probs=c(.025,.975))
quantile(post.df[,2],probs=c(.05,.95))
quantile(post.df[,2],probs=c(.10,.90))
tempdir()

Answer 1

Difficult to say for sure without sitting at your PC... Maybe it is something to do with R2WinBUGS looking in the wrong directory for WinBUGS.exe? You can point R2WinBUGS to the right place using the bugs.directory argument in the bugs function.

If not, try and install OpenBUGS and give R2OpenBUGS a go.