Hello,
This is an ongoing discussion I have had with colleagues and was wondering if I get some guidance from the MARK community. I have worked with several data sets on snakes where some individuals receive surgically implanted radio transmitters. Often, we either find a transmitter by itself or a snake doesn't emerge from the den in the spring. Usually, researchers (including myself) have assumed these represent moralities, yet we technically don't know. Occasionally a snake may expel a transmitter although mortality is more likely then expelling a transmitter. Hence the "tradition" of considering these probable mortalities as true mortalities.
But I am wondering about how these "probable" mortalities may affect (i.e., may bias) model-based estimates of survival. In any model where known-fates can be incorporated (e.g., known-fate, Burnham's joint recovery model), is it better to code "probable" mortalities as known mortalities, or does this risk under-estimating survival? Or can I even assume than wrongly assigning mortalities would under-estimate survival (maybe it doesn't with high recapture probabilities)? Would a more conservative approach be to censor such individuals after their last confirmed encounter?
Thanks,
Javan