using predict to detect interactions

[Chris Tyson]

Hi Chris,

I have automated radio tracking data with which I’d like to do some social network analyses. There are irregular gaps in the data for each individual though, which means there are not many simultaneous detections across all individuals. I could use some time threshold to group detections, but I also thought that using predict() with some set values for t would yield more simultaneous detections and make better use of the uncertainty in the tracking data. (Though given that the data is relatively noisy, I expect the periods between detections will indeed have a lot of uncertainty.)

Does using predict(), with some number of iterations, seem appropriate for my purpose? Any advice/recommendations are most appreciated!

Thanks,

Chris

[Christen Fleming]

Hi Chris,

Sorry for the delay. There is the distances() function that will predict pairwise distances and propagate those uncertainties. It uses predict() under the hood. From that you could calculate the probabilities of being within some distance over many times, and then get the predicted fraction of time spent within some distance. If there is a signal, this will get washed out with more uncertainty, so you might just use the sampled times where the pairwise distance predictions are the most accurate.

Best,

Chris

[Chris Tyson]

Hi Chris,

No worries, thank you very much for the response. I'll look into distances() and give that a try. 

Cheers,

Chris