joint model with few dead recoveries

[roodbergen]

Dear people,

The dataset I am analysing consists of 132 colour-ringed godwits, ringed during 3 years (2002-2004) and resighted during 3 years (2003-2006) in 4 study areas. 7 birds have been recovered dead, all in my study areas.
I have tried 3 different model types to analyse the data set:

1) recaptures only (knowledge on dead birds lost, resighting rate probably underestimated),
2) recaptures only with dead recoveries marked as loss at recapture (dead birds excluded from dataset after last resighting, local survival probably overestimated) and
3) Burnhams joint model for live resightings and dead recoveries (dead recoveries included as such, but estimate for reporting rate not precise, due to low number: CI=0.1-0.8 for one area)

Fortunately, the different model types give similar results. Both local survival and resighting rate are high and show the same pattern.
Phi's estimated with method 1 < phi's estimated with 3 < phi's estimated with 2.
I was hoping you could give me some advice on which of the three model types to use.
Thank you.

Best regards,

Maja Roodbergen