Estimate Absolute Number of Encounters

[Odd Jacobson]

Hi Chris,

I'm trying to estimate the total number of encounters between two animals during a shared observation period. I used the rates() function on their UDs to calculate pairwise encounter probabilities in units of 1/m^2. My initial idea was to multiply this by the intersection area between the UDs, but I'm uncertain if that's the correct approach.

According to the encounter statistics documentation (https://ctmm-initiative.github.io/ctmm/reference/encounter.html), it mentions that these values "must be multiplied by the square encounter radius (in meters) to obtain other values." This seems promising, but I'm not entirely sure. Specifically, I have two questions:

1) Will multiplying by the square encounter radius yield the absolute number of encounters in the observation period, and if so, how do I calculate this radius?

2) If this is not the correct method, do you have an alternative approach?

Thank you,

Odd

[Christen Fleming]

Hi Odd,

I just recently updated that function to the predicted encounter probability divided by the square encounter radius ( and changed the name to encounter(), with the CDE function name changed to cde() ). The immediate output is the expected proportion of time predicted to include 1-meter encounters, and if you multiply by 100^2 then you get the expected proportion of time predicted to include 100-meter encounters. The distance you plug in just needs to be much smaller than the home-range scales for the approximations used.

On the other hand, if you want the observed proportion of times, then you want the distances() function. But if the encounter probability is low, then you can often observe nothing.

Best,

Chris

[Odd Jacobson]

Hi Chris,

Thanks for the explanation. This is very helpful. Just to make sure I understand, when you say the proportion of "time" predicted to include 1-meter encounters, do you mean the proportion of the lifetime of the animals assuming a consistent underlying movement process? Alternatively, if I was interested in the proportion of time during the observation period predicted to include 1-meter encounters I would use distance()?

Also, just to make sure I understand "1-meter encounter" versus "100 meter encounter", is a 1-meter encounter when the movement paths are predicted to come within a 1 square meter proximity of each other at the same time? Then a 100-meter encounter is when they are within 100 square meters of each other at the same time?

Thank you,

Odd

--
You received this message because you are subscribed to a topic in the Google Groups "ctmm R user group" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/ctmm-user/4sW0VjpoOIg/unsubscribe.
To unsubscribe from this group and all its topics, send an email to ctmm-user+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ctmm-user/dc5052c1-eb2c-4305-8459-6fceecda4c58n%40googlegroups.com.

[Christen Fleming]

Hi Odd,

Yes and yes.

You plug in the radius (pairwise distance), so the area would be pi*r^2. We will have a paper out on the method soon.

Best,

Chris

[Odd Jacobson]

Hi Cris,

Thank you! Also, I noticed a new vignette on pairwise interactions (https://ctmm-initiative.github.io/ctmm/articles/interactions.html) - this is very helpful.

I have two questions after reviewing the vignette:

1)  How well do methods like distance(), proximity(), and encounter() perform when comparing animals sampled non-simultaneously? In my case, I am interested in studying interactions among 10 capuchin groups, each sampled over the same sampling window (over a calendar year), but not simultaneously. I am working with handheld GPS data (collected by human observers following habituated groups), therefore the data could only be collected on one group at a time. Does it still make sense to use these methods? Are the continuous-time movement models capable of accurately predicting interactions during unsampled times (gaps in data)? 

2) Also, with regards to a slightly different dataset, would you recommend these methods when the data is independent (~1 location per day)? 

Thanks,

Odd 

To view this discussion on the web visit https://groups.google.com/d/msgid/ctmm-user/0c0e2256-c283-4e56-b4ef-666739225676n%40googlegroups.com.

[Christen Fleming]

Hi Odd,

1. Yes and you get predictions with confidence intervals in the gaps. So, at times that are not well resolved, you will have larger uncertainties.
   
2. The confidence intervals will just widen out and you will get back nothing but uncertainty. It won't hurt anything, but you won't get anything.
   

Best,

Chris

[Odd Jacobson]

Thanks Chris!