If a function gives error try that iteration again (R) - r

Question

There are similar questions on StackOverflow, however, I am failing to find a solution identical (or remotely similar) to my issue. I am running a simulation as follows in R
iterSim <- 1000
for(i in 1:iterSim){
#Generate some random data
Z <- NormalDGP(n=T, beta=coef_min, theta=Theta, rho=Rho, ...)
y <- Z[1:(length(Z[,1])-1),1]
x <- Z[2:length(Z[,1]),2]
#Conduct the following tests! However, Dec_POS_Dep sometimes gives an error
Dec_POS_Dep <- POS_Dep(y, x, simul=TRUE, trueBeta=coef_min, ...)
Dec_POS_Fix <- POS_Fix(y, x, simul=TRUE, trueBeta=coef_min, ...)
Dec_CD_95 <- CD_95(y, x, simul=TRUE)
}
where for each iteration i random numbers are generated and three tests are run - i.e., Dec_POS_Dep,Dec_POS_Fix and Dec_CD_95. Unfortunately, sometimes in the simulation Dec_POS_Dep gives an error and the simulation terminates. I am not looking for the loop to skip an iteration if an error is given (as per many suggestions on StackOverflow); however, I would like that iteration to be repeated. E.g. If the code is on the 265th iteration and Dec_Pos_Dep gives an error, I want it to give it many more shots at the 265th iteration. Some solution to this would very much be appreciated.

Answer 1

Two things stand out as broken here:
Use of try (as MartinGal suggested) or tryCatch will allow things to continue. As far as rerunning that iteration, you'll need to keep track somehow of these failed runs and run them yourself, there's no notion of telling R to repeat a for loop iteration.
You are discarding data on each iteration, Dec_CD_95 is overwritten each time. Perhaps you mean to keep things around?
Here's a suggestion:
iterSim <- 1000
out <- list()
while (length(out) < iterSim) {
try({
#Generate some random data
Z <- NormalDGP(n=T, beta=coef_min, theta=Theta, rho=Rho, ...)
y <- Z[1:(length(Z[,1])-1),1]
x <- Z[2:length(Z[,1]),2]
#Conduct the following tests! However, Dec_POS_Dep sometimes gives an error
Dec_POS_Dep <- POS_Dep(y, x, simul=TRUE, trueBeta=coef_min, ...)
Dec_POS_Fix <- POS_Fix(y, x, simul=TRUE, trueBeta=coef_min, ...)
Dec_CD_95 <- CD_95(y, x, simul=TRUE)
out <- c(out, list(Dec_POS_Dep, Dec_POS_Fix, Dec_CD_95))
}, silent = TRUE)
}
This is a little sloppy, admittedly, but it should always end up with 1000 iterations of your simulation.