simulation of an inhomogeneous poisson process

Hello everybody
I've been working on a model with an inhomogeneous arrival process. For this I created a code for an inhomogeneous poisson arrival. It works in the R script, but when I use a similar code in flexsim HC, the results are not the same. Attached are the test model and code of the R scripts. Is somebody able to explain the diffrence? Error/diffrence in my flexsim code?

model:

rm(list=ls())

DPfad <- "C:\Users\" # the path of the method arrival
setwd(DPfad)
source("arrival.R")

t=0
#lambda=runif(24)
lambda = 1:6
lambda = rep(lambda,4)

event=c()
for(i in 1:1e5)
{
iat=arrival(t,lambda)

t=t+iat
event=c(event,t)
}

hist(event%%24,breaks=0:24,freq=FALSE)
points(0:23+0.5,lambda/sum(lambda))

# comparison with flexsim data
## load data
sim <- read.csv(paste(DPfad,"test_model2.csv", sep=""), sep=",")

# transform minutes into hours
sim$TIME <- sim$TIME/60

# create a histogram
hist(sim$TIME%%24,breaks=0:24,freq=FALSE)
points(0:23+0.5,lambda/sum(lambda))

#------------------------------------------
arrival:

# Funktion zur Berechnung der nächsten Interarrival Time
arrival=function(t,lambda)
{
t0=t # aktuellen Zeitpunkt speichern
p=runif(1) # Zufallszahl ziehen
F=0; # Initialisierung der Verteilungsfunktion
l=lambda[trunc(t%%24)+1] # aktuelles lambda ermitteln
dt=ceiling(t)-t # Zeitspanne bis zum nächsten lambda
P=(1-exp(-l*dt)) # Dichtefunktion der aktuellen Zeitspanne

while(F+(1-F)*P {
F=F+(1-F)*P # Verteilungsfunktion erweitern
t=t+dt # Zeitspanne hinzufügen
l=lambda[trunc(t%%24)+1] # aktuelles lambda berechnen
dt=1 # aktuelle Zeitspanne
P=(1-exp(-l*dt)) # Dichtefunktion der aktuellen Zeitspanne
}

t=t-log((1-p)/(1-F))/l # letzte angebrochene Zeitspanne hinzufügen

return(t-t0) # berechneter Zeitunterschied
}

Thanks for your respond.

If you're suspecting a mismatch in code translation, have you tried putting breakpoints into your code to check the value of variables at various steps in the code? Go to Tools>Debugging>Breakpoints to turn on breakpoints, and then simply click to the left of each line in the interarrival time code window to add a breakpoint (red dot will appear). As you run the model, the breakpoint manager window will popup allowing you to further investigate your code.

There may be a much easier and less cumbersome way to model the nonhomogeneous arrival process you're dealing with using a Log-Linear function. I'm assuming, of course, that your underlying pattern is Poisson distributed. Whatever the case, if you'd send me your data, I might be able to create the arrival pattern you're looking for.

Thanks,

Lou Keller

Originally Posted by Lou Keller

There may be a much easier and less cumbersome way to model the nonhomogeneous arrival process you're dealing with using a Log-Linear function. I'm assuming, of course, that your underlying pattern is Poisson distributed. Whatever the case, if you'd send me your data, I might be able to create the arrival pattern you're looking for.

Thanks,

Lou Keller

hi i am a student in tp and doing a project for a hospital by using the flexsim hC, i do not know how to use the expertfit to create the arrival patten . i have the visiting data, could u kindly help me? thx