Message-ID: <5176B363.1020206@tu-dresden.de>
Date: 2013-04-23T16:14:27Z
From: Thomas Petzoldt
Subject: [R-sig-dyn-mod] metapopulation SIR model
In-Reply-To: <1366728364.81874.YahooMailNeo@web171503.mail.ir2.yahoo.com>
Dear Vincent,
I appreciate that you are fighting with the topic, but not all modelling
questions can be solved by volunteer support. As far as I see, your
fluxrate approach seems to go into a promising direction, but don't
have time to check if "it is all right".
On 23.04.2013 16:46, vincent laperriere wrote:
>
> I tried out to adapt the n-species LV model as you mention in your
> last message but did not succeed. Is it possible to specify a model
> as:
>
> model <- function(t, S, I, R, parms) {......}
This is not possible because the deSolve help file ?ode tells us:
"If func is an R-function, it must be defined as:
func <- function(t, y, parms,...)."
> and then run it as:
>
> out <- ode(S0, I0, R0, times, model, parms)
No, it would not follow the syntax of R/deSolve's "ode" function. States
have to be bound together as a vector. But, you have the option to
reorganize your states internally in the model function, as long as
states at the beginning and derivatives at the end remain a vector and
occur in the same order.
> Nevertheless, I tried another way to develop a SIR metapopulation
> model, like this: do you think it is all right with the commands
> below?
See above, sorry. Nevertheless, I've put some code together yesterday
evening, just for recreation ;-) Maybe this gives you some idea:
http://hhbio.wasser.tu-dresden.de/projects/modlim/src/SIR-metapop.R
I'm sure this can be made even simpler, more realistic or whatever.
Maybe that someone else on this list can give you more insight ?!!!
Comments are welcome.
thpe