by abreton » Wed Apr 26, 2006 6:00 pm
When trying to decide on a solution to your problem, remember that the the CJS model provides estimates of apparent survival over the intervals that occur between sampling occasions - i your study, these intervals would include what you described as "several months downtime during the winter until I pick back up again in the spring". If this is clear, then you need to focus on one important assumption of the CJS and other CMR models: sampling is instantaneous. Although it is not possible in wildlife, or perhaps any other CMR setting, to instantaneously sample/resample, what is important, for this assumption, is that none (or very few) animals are lost during the sampling period, i.e., no mortality occurs over the sampling interval. If, e.g., I resight marked sea turtles on a beach over 1 week, my resampling period is certainly not 'instantaneous'. However, as long as none (ideally) or few animals are lost during this week, then the instantaneous assumption (IA) is not an issue. But say, unknown to you, 20 out of 100 turtles die during the week after being caught in fishing gear - this would certainly be a strong violation of the IA and it demonstrates why the assumption is 'biologically' important: if you ignored the fact that 20% died during the sampling period and calculated survival probabilities between sampling occasions, then your estimates of survival over intervals is confounded by survival over sampling periods. And most likely, you'd be unaware of the losses over the sampling period and any inference you make would be a long way from reality.
So you need to shorten your resampling interval to something that satisfies the IA - perhaps 2 weeks to 1 month maximum depending on how much you can stretch the IA given the life history oy your study species - you'll have to 'dump' any recaptures that fall outside of your sampling intervals. Once you reduce your resampling interval to a biologically reasonable time span given the study species, then you'll have three sampling events or intervals and two survival events or intervals: 1st Sampling interval - 1st Survival Event - 2nd Sampling Interval - 2nd Survival Event - 3rd Sampling Interval. Notice that you'll be estimating survival from one sampling interval to the next and that the survival interval will encompass most of the year.
Alternatively, you could collaspe your sampling intervals to two (or more)/year - as you suggestd in your post. If you go this route, you'll have more sampling events and thus, more intervals over which to estimate apparent survival. Ultimately, if you want to use CMR models, you'll need to reduce the duration of your sampling intervals. I've had some experience with this and have a paper 'nearly' accepted in TAFS on slimy sculpin (New Brunswick) - if you want a copy send me an email. Good luck.