I am solving numerically stochastic differential equation with use of diffeqr in R, check if the last value is large enough and replicate the process for 5 times. The piece of code for this part is:
for(i in 1:30){
for(j in 1:15) {
u0<-c(1.5e 6 i*5e 4,j*2e 5)
data<-replicate(n=5,{
prob <- de$SDEProblem(f,g,u0,tspan,p)
fastprob<-diffeqr::jitoptimize_sde(de,prob)
sol <- de$solve(fastprob,saveat=0.005)
mat <- sapply(sol$u,identity)
udf <- as.data.frame(t(mat))
udf[40000,1]<1e 6}
)
sum(data)/length(data)
res02_4[i,j]=sum(data)/length(data)
}
}
However, sometimes of the the results is NA for whatever reason and so sum(data)/length(data)
is also NA. In this case I want to redo entire replicate
piece (until it is not NA). How I can organize this? First thing comes to mind is a combination of if() {} else {}
but I think it'll be quite cumbersome construction - is there more elegant way?
Thank you in advance!
CodePudding user response:
You could pre-define data
as NA
, then wrap the replicate
section in a while
loop that runs one or more times until sum(data)
is not NA
for(i in 1:30){
for(j in 1:15) {
u0<-c(1.5e 6 i*5e 4,j*2e 5)
data <- NA
while(is.na(sum(data))) {
data <- replicate(n = 5,{
prob <- de$SDEProblem(f,g,u0,tspan,p)
fastprob<-diffeqr::jitoptimize_sde(de,prob)
sol <- de$solve(fastprob,saveat=0.005)
mat <- sapply(sol$u,identity)
udf <- as.data.frame(t(mat))
udf[40000,1]<1e 6}
)
}
sum(data)/length(data)
res02_4[i,j]=sum(data)/length(data)
}
}
Obviously, I can't demonstrate that this works because you haven't included the de
object in your question.