[R] Stochastic SEIR model

Massimo Fenati massimo.fenati at infs.it
Thu Nov 16 16:27:08 CET 2006


Thanks for your fast advises.

A simple examples of SEIR model is shown below:

#expample of very simple SEIR model#####
library(odesolve)
times<-seq(0,1200,1)
parms<-c(
	b=0.35,		#BETA OR COEFFICIENT OF TRANSMISSION
	pl=1/7,		#LATENCY
	g=1/21,		#RECOVERY
	a=0.05/7	#LETHALITY IN ADULT
	)
xstart<-c(S=100,E=0,I=1,R=0,nn=101)	
model<-function(t,x,parms){
S	<-x[1]
E	<-x[2]
I	<-x[3]
R	<-x[4]
nn	<-x[5]

with(as.list(parms),{
dS<--S*b*I/nn
dE<-S*b*I/nn-E*pl
dI<-E*pl-I*(a+g)
dR<-I*g
dnn<-dS+dE+dI+dR

list(c(dS,dE,dI,dR,dnn))
})
}
out<-as.data.frame(lsoda(xstart,times,model,parms))
plot(times,out$S,type="l")
lines(times,out$E,col="green")
lines(times,out$I,col="red")
lines(times,out$R,col="blue")
#########

I'd like to vary one or more parameters (with several 
distribution) for obtaining probabilistic results from the 
model projections.

Now I'll try also with winBUGS, but I've never worked with 
it. I hope..

Thank you very much.

Max



On Thu, 16 Nov 2006 09:26:12 -0500
  Tamas K Papp <tpapp a Princeton.EDU> wrote:
> On Thu, Nov 16, 2006 at 02:55:07PM +0100, Massimo Fenati 
>wrote:
>> Dear colleagues,
>> I’m a new R-help user. I’ve read the advertisements 
>>about 
>> the good manners and I hope to propose a good question.
>> I’m using R to build an epidemiological SEIR model based 
>> on ODEs. The odesolve package is very useful to solve 
>> deterministic ODE systems but I’d like to perform a 
>> stochastic simulation based on Markov chain Montecarlo 
>> methods. I don’t know which packages could be used to do 
>> it (I tried with "sde" but without results).
>> Have you some suggestions about useful methods and/or 
>> function in R for reaching my aim.
> 
> Hi Max,
> 
> I don't know what SEIR is.  R has some MCMC packages
> (RSiteSearch("MCMC") will help you there), but it is 
>easy to write a
> Metropolis/Gibbs sampler yourself, making use of the 
>structure of your
> problem.
> 
> The sde package is for approximating likelihoods of 
>continuous time
> stochastic differential equations (which is a difficult 
>problem by
> itself).
> 
> It is hard to give more advice without knowing what you 
>are trying to
> achieve -- it is good that you have read the posting 
>guide, but please
> give more details if you want more specific answers.
> 
> HTH,
> 
> Tamas

MASSIMO FENATI
-----------------------------------------
Istituto Nazionale per la Fauna Selvatica
Via Cà Fornacetta, 9
40064 - Ozzano dell'Emilia (BO)- Italy
tel: +39 0516512245
cel: +39 3392114911
fax: +39 051796628
e-mail: massimo.fenati a infs.it



More information about the R-help mailing list