Complete detection matrix

42 views
Skip to first unread message

César Llanos

unread,
Aug 23, 2022, 8:55:25 AMAug 23
to unmarked
Hello,

I'm a biology student, so I'm sorry if I'm asking something very simple. I have not found the answer in other threads.
I am using dynamic models, but what I am not able to get is the full detection probability matrix, with a value for each sampling day. I do get the detection trend over the years, but I would like to see it for each sampling day.
Would you know which function to use?

Thanks in advance

Hardin Waddle

unread,
Aug 23, 2022, 9:04:02 AMAug 23
to unma...@googlegroups.com
Hi Cesar-

I think what you need to do is set an observation covariate (obsCovs) for sampling day. I just recently did an analysis like that for turtles where we think they are much more detectable on the first day of a three day trapping session than the second and third day. I have a matrix called turtle of 15 sites with three visits to each. The following code works to let me estimate detection on each day:

visit <- matrix(as.character(1:3), 15, 3, byrow=TRUE)
turtleUMF <- unmarkedFrameOccu(y=turtle, obsCovs=list(visit=visit))

This is a single-season model, but it should work the same way for a dynamic model. It might be more complicated if you wanted to have each primary period have separate estimates of detection for each sampling day.

Hardin Waddle

--
*** Three hierarchical modeling email lists ***
(1) unmarked (this list): for questions specific to the R package unmarked
(2) SCR: for design and Bayesian or non-bayesian analysis of spatial capture-recapture
(3) HMecology: for everything else, especially material covered in the books by Royle & Dorazio (2008), Kéry & Schaub (2012), Kéry & Royle (2016, 2021) and Schaub & Kéry (2022)
---
You received this message because you are subscribed to the Google Groups "unmarked" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unmarked+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/unmarked/f1d56692-39f9-4dd6-8ab7-479d332c86dfn%40googlegroups.com.

César Llanos

unread,
Aug 24, 2022, 5:12:16 AMAug 24
to unmarked
Hello, Hardin.

Thanks for the advice. Still, my intention is precisely to have each primary period with separate estimates of detection for each sampling day.
Any ideas for this?

Hardin Waddle

unread,
Aug 24, 2022, 8:50:28 AMAug 24
to unma...@googlegroups.com
Well I think it depends on what exactly your hypothesis is.

In our case, we had reason to believe that during a trapping session the detection decreases from day 1 to day 3. Thus we only cared about the day of the trapping session.

If you think every day could be different and want to test that, then you could do the same thing I showed, but have a unique factor value for each day. You would need to figure out if that hypothesis makes sense to you.

An alternative that you might consider is thinking of covariates that would predict detection like temperature, precipitation, or day of the year (Julian day). Those might be easier to interpret than a unique value for each day of sampling.



César Llanos

unread,
Aug 25, 2022, 4:14:57 AMAug 25
to unmarked
Ok, thank you!

And, what about getting the detection probability in each season and each place? Could it be possible? I'm not sure what function could I use in that case...
In doing so, I would like to compare relative abundance data with the detection probability.

Thanks in advance

Hardin Waddle

unread,
Aug 25, 2022, 8:52:43 AMAug 25
to unma...@googlegroups.com
I think if you attempt to estimate a separate detection estimate for each sampling occasion at each site, you are going to have a model with too many parameters. What would work better is to identify covariates of detection that vary across sites and that you think would have an effect on detection, and model those.


César Llanos

unread,
Aug 28, 2022, 3:20:44 AMAug 28
to unmarked
Thank you very much for the advice, Hardin.

Finally, what I need is a table with detection probability values at each location and season.
I don't need the probability of each survey.
Would you know how to do it?
Reply all
Reply to author
Forward
0 new messages