Questions about Lukac's Young model

[Kcolson]

This is my first post, so I apologize if it's in the wrong sub-forum. I'm looking to use Lukac's young survival model from radio marked adults on a set of observations to examine time specific survival and non-detection in moose calves (where we have an appreciable number of twins). However, in the 2004 publication, it stated that the initial observations should not be lower than subsequent observations. I'm concerned about this, as on the first survey we may miss one of the young. Since both calves are born at the same time, it's improbable that the moose has gained a calf in that time.

If anyone has any insight as to whether this effects the model as implemented in MARK, I would be greatly appreciative. Ideally, these sorts of events are exactly what we wish to draw data from.

My second question is more methodological - Should data, when entered into MARK, be 'left aligned?' That is to say, the first sampling interval for an individual is the first that it is observed with young? Or should intervals be aligned by survey date? My off the cuff intuition is that the answer depends on whether I'm more interested in time specific or age specific survival, but I'm curious about you're all's response.

Thanks for your time!

[Paul Lukacs]

The model conditions on the number of young alive at the start of sampling. For the original development, we were counting eggs in nests so that was easy. If you don't know the original number of young, then detection and survival are not separable. If you can quickly determine with a high degree of certainty whether a cow has 1 or 2 calves, then you could start your analysis at that point in time rather than at birth. You may miss some initial mortality, but should have good results after that. If it is difficult to be certain about whether a cow has one or two calves, then survival and detection cannot be separated as it's always possible that you're detecting a different calf.

If you simply assign a 2 at birth for any cow subsequently detected with 2 calves, you will bias survival high. The magnitude will depend on detection probability and survival rate.

You can format your data either left aligned or aligned by time with leading zeros. I just depends how you want to interpret time (left aligned time=age).

Paul

[Kcolson]

Hi Paul, thanks for your speedy reply.

Since estimating detectability is of interest in our system, wouldn't censoring our leading non-detection events bias our detectability high? Or are you instead suggesting only looking at our entire data set at some point past when calf status can be accurately ascertained at least once (e.g., after the 4th interval, when we have got an accurate count on all young at least once)?

Best,
Kevin