2

I have the following codes to run. 

library(fExtremes)
library(maxLik)
theta0 <- c(0.1, 1)
theta1 <- c(0.1, 2)

set.seed(20150518)

X <- matrix(0, nrow=1000, ncol=1000)

for(i in 1:1000){
  X[i, ] <- c(rgpd(900, xi=-theta0[1], beta=theta0[2]), 
              rgpd(100, xi=-theta1[1], beta=theta1[2]))
}

library(maxLik)

loglik <- function(param){
  shape <- param[1]
  scale <- param[2]
  sum(dgpd(x, xi=-shape, beta=scale, log=TRUE))
}

scale.mle <- rep(0, 1000)

for(i in 1:1000){
  x <- X[i, ]
  scale.mle[i] <- as.numeric(maxLik(logLik=loglik, start=c(theta0[1], theta0[2]), fixed=1)$estimate[2])
}

However, for certain runs I get the following error message:

Error in maxNRCompute(fn = logLikAttr, fnOrig = fn, gradOrig = grad, hessOrig = hess,  : 
  NA in the initial gradient

And what is worse is that once this error occurs, the entire for loop is stopped even if later data input can be used. How can I ask R to skip such problematic case and move on to the next. In order words, how do I continue to finish all the 1000 runs without stopping. Thanks.

Update: Thanks for pointing out the other question. However, I still do not get it. Could anyone illustrate the usage of tryCatch with my example, please?

LaTeXFan
  • 1,136
  • 4
  • 14
  • 36
  • @josilber Could you illustrate the usage of `tryCatch` using my example, please? I still do not get it. Thanks. – LaTeXFan Oct 21 '15 at 00:28
  • 1
    Right now you have `scale.mle[i] <- as.numeric(...)` ; you would replace that with something like `tryCatch({scale.mle[i] <- as.numeric(...) }, error = function(e) { scale.mle[i] <- NA })`. The linked question has a pretty good answer, so I'd encourage you to read that for more details. – josliber Oct 21 '15 at 00:30

0 Answers0