Options for ragged telemetry data
Posted: Thu Feb 24, 2011 2:09 pmI am looking for suggestions regarding the best method to use to model the data described below:
I have two years of data on two groups of birds: translocated and resident. The first year of data contains continuous daily survival information on the residents. The first year of translocated bird data is only at a weekly frequency and is extremely spotty for the first 4 or 5 weeks of tracking. The second year of data for both groups contains continuous and daily observations.
I do not want to remove the first 1.5 month of data to avoid the ragged translocated telemetry data from the first year, as most of what I am trying to do is to test for a difference between translocated and resident survival and that is a relatively important time period. Also, the gaps in the data are not during the same time period for each bird, so it does not seem practical to redefine time intervals in a known-fate model either (ie during a period of three weeks one or two birds may not have been looked for and all the others were). I have considered the use of multistrata models as used in a recently published lynx paper, however, I was hoping to use those to estimate habitat specific and not state specific survival and movement. Using those models would also only provide me with weekly survival estimates and I was hoping for a finer resolution (if that's even possible when we didn't collect the data daily).
The best option I have been able to cobble together was the use of the nest survival procedure in program MARK for the first year of data (for both residents and translocated birds) and then known fate models for the second year of data. It seems that comparison across years would be difficult if this route were followed though as the models couldn't be compared (?)...
I would appreciate any suggestions regarding a model type that would optimize the comparison between translocated and resident birds as well as a comparison across years for these groups.
I have two years of data on two groups of birds: translocated and resident. The first year of data contains continuous daily survival information on the residents. The first year of translocated bird data is only at a weekly frequency and is extremely spotty for the first 4 or 5 weeks of tracking. The second year of data for both groups contains continuous and daily observations.
I do not want to remove the first 1.5 month of data to avoid the ragged translocated telemetry data from the first year, as most of what I am trying to do is to test for a difference between translocated and resident survival and that is a relatively important time period. Also, the gaps in the data are not during the same time period for each bird, so it does not seem practical to redefine time intervals in a known-fate model either (ie during a period of three weeks one or two birds may not have been looked for and all the others were). I have considered the use of multistrata models as used in a recently published lynx paper, however, I was hoping to use those to estimate habitat specific and not state specific survival and movement. Using those models would also only provide me with weekly survival estimates and I was hoping for a finer resolution (if that's even possible when we didn't collect the data daily).
The best option I have been able to cobble together was the use of the nest survival procedure in program MARK for the first year of data (for both residents and translocated birds) and then known fate models for the second year of data. It seems that comparison across years would be difficult if this route were followed though as the models couldn't be compared (?)...
I would appreciate any suggestions regarding a model type that would optimize the comparison between translocated and resident birds as well as a comparison across years for these groups.