multi-state model and "true" recruitment

[Claudio]

Hi all,
I am analysing CMR data from an amphibian population studied for four years. Individuals were captured as juveniles or adults. Juveniles stage span one year, that is in the year following the first capture a juvenile or permanently emigrates or becomes an adult. I am especially interested in recruitment, and I thought that a multi-state model will have allowed to me to disentangle juvenile survival and the attainment of sexual maturity (i.e. recruitment).
Thus, I tought up to use a multistate model, with state J(uvenile) and A(dult), fixing S_J=0, p_J=0, psi_A-->J=0; for psi_J-->A I built a cohort PIM and fix psi for the second and third interval =1. Actually I have 8 models, with S_A and p_A constant or time variable, and psi_J-->A costant or variable.
The point is that I always obtain psi_J-->A=0.45 with S.E.=0. After playing around (stepwise fixing of paramenters), I discovered that this is due S_J=0. (by the way, 0.45 is close to ther ratio [juveniles captured as adults / total amount of juveniles])
I fix S_J=0 because no juvenile captured as juvenile will be recaptured as juvenile in the year following: it will go out of the population or it will become an adult. Thus, my plan was that psi_J-->A would have told me the rate of "true" recruitment, and (1 - psi_J-->A) the juvenile mortality. But something is wrong: what?
thanks in advance
Claudio

[jlaake]

You are confusing Psi and S. Setting S_J=0 will prevent any juveniles from surviving and then transitioning to Adult. 0.45 is the value you'll get when there is no information for the parameter and it has nothing to with the ratio. What you are thinking about is that Psi J to J is 0 and not S_J=0. Now I'm not sure you are thinking about this clearly because if you are using annual intervals and you set Psi J to J=0 that means Psi J to A=1 so I'm not sure what you are thinking you are measuring as "recruitment".

--jeff

[Claudio]

Thanks for your reply.
Now I have understood that what I was doing did not make much sense.
Now I am thinking to model S_J basing on cohorts and fixing psi_J = 1 starting from the second year.
thanks again
Claudio