incorporating lifespan into multi-strata models

[arpat]

Hi!

I have been using the multi-state model to analyze the survival and maturation (transition) rates of non-reproductive and reproductive animals. One main problem is that I have too many occasions (~100) and it takes too long for a single model to run; it averages up to 6-7 hours for each model. I am planning to run a series of models on this dataset, and I am trying to find ways of shortening the model running time.

The number of occasions that the study spans (~100 occasions) is much higher than the natural lifespan of the individuals (~10 occasions). I believe that this causes the main delay. For example, if we consider an individual with encounter history:

0010122022000000000000000000000000000000000000

(where 1=non-rep, 2=rep)

program MARK does not know the natural lifespan of the individuals. If I am not mistaken, it assumes that an individual can survive till the end of the study period and tries to calculate all the possible fates of an animal for the period after the animal was last seen (10th occasion). In reality, however, the animal is known to be dead for sure after the 13th occasion. I wonder if there is a way of incorporating the natural lifespan of the animal into the models, and if this makes the model calculations easier and faster.

I was thinking of coding the occasions in which the animal is sure to be death as a separate state (e.g. X). So my recapture history becomes:

001012202200XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

(where 1=non-rep, 2=rep, X=known to be dead)

I was planning on setting both the survival and recapture rates within the X state equal to 1, and transition from X state to other states (1&2) equal to 0. This way I was hoping to speed up the model running process.

Another option, that I can think of is using multi-strata models with live and dead encounters, and putting a hard cap on the animals encounter history by assigning a dead encounter at the end of the lifespan. However, these models have even more parameters, and I am not sure if it will really speed up the process.

I am not too familiar with the underlying structure of multi-strata models, and I might be making a major mistake while trying to speed up the process. I will be grateful if someone can comment on the above approaches. I will also be grateful if anyone could give me some other suggestions on how to speed up the model running process in this case.

Thanks a lot,

Arpat Ozgul

[arpat]

Ooops..

I see that "001012202200XXXXXXX.." approach obviously messes all the survival rates of stages 1 and 2, and also other parameters. So it will definetely not work.