Hello,
We are conducting a SMR study and about half of our capture events were photographed at night in infrared. We are identifying individuals via colored duct tape on GPS collars. As such, our ability to identify individuals in IR photos is drastically different when compared to our day photos. Depending on the season and aid of telemetry, our ability to ID individuals during the day is about 47% to 84% whereas our ability to ID individuals in IR photos is about 3% to 48%. Since our IR photos constitute about half of our data, we would like to keep them in our analysis and account for the difference in ID.
To address this we have split our data into a multi-session model, where we have 1 session for day photos and 1 for IR. We therefore have secr.fit(CH, ...pID = ~session, ...). Our data is not independent and therefore, we bootstrapped to acquire larger confidence intervals. That said, we are having issues with the bootstrap producing values that surround our point estimate and are now revisiting our options. Regardless of our bootstrap issue, do you suggest a better method to account for our ability to ID the IR photos?
Our alternative idea is to split each day into 2 occasions, where day 1 = occasion 1, night 1 = occasion 2, etc.
I am also using secr 4.3.3 on Windows OS.
Any suggestions would be appreciated.
Thank you