Nest-survival model encounter history

[Kent]

Hi Everyone-
I had a question about data coding that has bothered me.

I am stuck on the fact that if you use the last day checked and alive as the end date of a nest, then are you not biasing survivorship estimates low because that was not the entire nest period? You missed n number of days the chicks were actually exposed from last date know alive to last nest check (where you found it empty and coded as fledged for any number of reasons).

Perhaps, part of the problem is that this method was developed with the assumption that hatch date was known exactly (or could be predicted based on known nest age). Everything subsequently has also carried this assumption. But, here and many other studies, we wish to use it for nestling stage.

Here is a statement in a paper in Studies of Avian Biology 34. On page 119 to 120 they say, "For successful nests, an attempt was made to estimate the actual day that the nest fledged young, rather than simply using the last day checked. The latter, if different, would unjustifiably add survival days to a nest when failure was no longer possible. You don't want to bias upward. But, when they say they attempted to ESTIMATE the actual day the nest fledged, that makes me think they did something rather than just using the last day known alive as the end (Took midpoint, estimated from age of young, estimated from median nestling days for species or something..... ).

So what is the correct way to code nests? For successful nests do you use the last day checked and known alive as the end of the nest? Do you estimate a day between the last day checked and alive and the last day checked? Is this method simply no good for altricial bird nests?

Thanks for any help.
Kent

[Jochen]

Hi!
I recommend to follow the suggestions in "the book", which deals with this problem, too. If you can be SURE that and when a nest was successful you may take the date of fledging as the last day. If you are NOT sure about the nest's fate then take the last day active (checked and known alive) as the last day and code the nest successful.

Applying this to the nestling stage means, that it is easy to know the likely fledging date, but that it may be quite difficult to be sure about the fate of a nest which was not checked at this time (see J. Field Ornithol. 77(4):357–364, 2006)

[abreton]

"So what is the correct way to code nests? For successful nests do you use the last day checked and known alive as the end of the nest? Do you estimate a day between the last day checked and alive and the last day checked? Is this method simply no good for altricial bird nests?"

The 'ideal' way is to code them in such a way that the results are most appropriate for your research question. If your question is about survival of nests from hatch to fledge, then the ideal way to code these includes for successful nests "the day that the nest attempt [was] successfully completed" (pg. 18-5, the book). If these data are not available, then additional assumptions will have to be made and inference adjusted accordingly. In your case, you'll likely have to replace "day successfully completed" in your nest survival histories with "last date known to be alive". And note that "last date known to be alive" is already part of the required data for these histories (page 18-5). Ultimately, this approach is not incorrect, it just isn't ideal if what you want to model and estimate is the daily survival rate from hatch to fledge. Your data are truncated with respect to the hatch to fledge period so you'll have to adjust your inference to account for this shortfall.

The method works well for both altricial and precocial species - but to acquire 'the most' from the method a minimum amount of information about the process is required including "the day that the nest attempt [was] successfully completed" for successful nests. Sounds like your study falls a bit short of this ideal but a lot can still be gained from applying nest survival models.

[abreton]

Making assumptions from the existing data to predict the day young fledged from successful nests is of course another 'less than ideal' option. And there may be others, I wasn't intending in my last post to provide a comprehensive list. The best option is a function of the available data and target research question.