www.phidot.org

This is a quick question for Murray that I decided to post here for archiving purposes. I am working with a hair snare dataset using 'proximity' detectors. However, the dataset does contain some capture histories where an animal was detected >1 time at a single site during a single occasion which suggests using a count-based detection model rather than a binary-based detection model. The occurence of these "multiple detections" is not frequent enough (in my opinion) to justify using a 'count' detector. My question is if I specify 'proximity' as the detector type in my call to the read.capthist function, will secr automatically (and more importantly, correctly) collapse those "multiple detections" into occasion-site specific binary detections or do I need to sift through the dataset and remove the redundant occasion-site specific detections?

Thanks
Jared

I believe secr.fit handles these appropriately (i.e. any number of detections will be treated as a '1' rather than a '0', though of course you have to override the data check with verify = FALSE). If in doubt, use reduce.capthist() like this to make the conversion:

Code: Select all

 myproxCH <- reduce(mycountCH, output = 'proximity')

Murray

Thanks Murray. That was very helpful. I have another question. For example, say I wanted to model annual density for a 3 year/2 group dataset as 'a,a,a,b,b,b' and model g0 as 'a,b,a,b,a,b' when the data is structured as 6 sessions. How can I specify unique session covariate arguments for different model parameters? Can the sessioncov argument be a list of factor vectors and then use specific elements from that list in the model statements for different parameters?

Thanks again
Jared

Hi Jared
On the help page for secr.fit, sessioncov is described as an 'optional dataframe of values of session-specific covariate(s)'. A dataframe is a special sort of list and, yes, you can include any number of different factors or other predictors in sessioncov and use any subset in the formula for any parameter.
Murray