Hello Phi.dot,

There are residents and transients in my population. I am using the ‘time-since-marking’ approach (MARK manual, ch. 7) with CJS models to handle the heterogeneity caused by transients, and to estimate the proportion of transients in the population. I have two groups, males and females. Below is an example of the parameter index matrix used to model phi with different survival probabilities for transients and residents (6 occasions, no time dependence):

1 2 2 2 2

1 2 2 2

1 2 2

1 2

1

This model gives two survival estimates, phi1 for newly caught animals (transients), and phi2 for those caught over multiple years (residents). Typically, phi1 should be lower than phi2.

When I run this model, the results for males make sense, with phi1 < phi2, indicating the prevalence of transients in the male population.

But in the results for females, phi1 (0.88) is HIGHER than phi2 (0.78), which doesn’t make sense (phi1 should always be lower I think). I have rechecked the capture histories and PIMs and cannot find any mistakes. All the beta and real estimates seem to be estimated ok (no crazy SEs, etc). I do have a small sample size, with just 12 females and 21 males. Running separate analyses on each sex (as opposed to a single analysis with 2 groups) makes no difference, I get the same results.

It is true that transience in females is less common than with males, and when I run another model with a transient effect for males only, and not for females (i.e., phi(.) for females), the model has substantially more AIC weight. But still, there were some female transients (individuals F6 and F7, in the capture histories below), so phi1 should be lower than phi2.

Is it valid that phi1 > phi2? Or does this indicate a mistake somewhere, or an artifact of small sample size?

Thank you for your help, Robert

F1 111110

F3 111000

F4 011100

F5 011111

F6 001000

F7 001000

F8 000110

F9 000110

F10 000011

F13 000001

F14 000001

F19 000001