Royle Count data analysis support

[PAUDEL]

Dear Members,

I am early user for the PRESENCE SOFTWARE. Using repeated count data i need to analysis following models:

Data: two seasons with three replication
Survey approach: Transect survey (count for each transect)

I need to do following analysis:

1. N estimates for each transect with 95% CI
2. N-abundance model using abundance data with co-variate depth, width, habitat type (categorical), season( categorical), disturbance (categorical) and water current (categorical)

3. Occupancy model using Presence/Absence data with above mentioned variables.

If anyone help me for this analysis it would be great help. I am early user, therefore i am seeking basic procedure for this from data management to model development as a quick book.

Thank you
shambhu paudel

[jhines]

Hi Shambhu,

The overview in the PRESENCE help menu is the best place to start. It gives detailed instructions on how to enter data into the program and run some basic models. Although I don't think it has specific instructions for building the Royal repeated count models, it does have instructions for other models with covariates and the procedure would be the same for the count models. There are also step-by-step tutorials in the help menu.

The most in-depth documentation for running occupancy or capture-recapture analyses is the 'mark book', which is available on this site.

Cheers,

Jim

[Bdbio]

Hello everyone. First of all,thank you very much for your participation in this forum, I'm also a researcher who has recently started to use PRESENCE in my analysis and I admit that your comments are really helpful.
Like PAUDEL said in the first comment, I also want to use Royle Count model to get an abundance estimation using covariates, using presence-absence data, but when I try to look for papers to see if someone has tried and published it before, I can't find anything.
I have studied all the statistical background of the analysis, and, despite the fact that, in my opinion, it is a very powerful tool, maybe it is hard to believe that from presence-absence data (actually, detection-non-detection data) one can get the abundance of the study area.
So, could it be considered a good approach? and
Why, using PRESENCE, it is not possible to assess the model by bootstrapping or Chi-Square?
Thank you very much.

[jhines]

Like you noted, it's hard to believe you can get abundance estimates from occupancy data, but if you look at how the detection probabilities are modeled in the Nichols-Royal paper, you can see how it's done. However, there are some pretty strong assumptions that must be made in order for this model to work. I think that might be the reason you haven't seen many papers which use the method.

The goodness-of-fit test can be 'flaky', even with the most simple models, since cell probabilities tend to be very small and one unusual detection history can cause a huge change in the test statistic. Adding a GOF test to other models is on my list of things to do, but it will take quite a bit of time to test. I suspect the abundance models will be even 'flakier'

[cooch]

Like you noted, it's hard to believe you can get abundance estimates from occupancy data, but if you look at how the detection probabilities are modeled in the Nichols-Royal paper, you can see how it's done. However, there are some pretty strong assumptions that must be made in order for this model to work. I think that might be the reason you haven't seen many papers which use the method.

Ditto. Part of the reason that Gary White resisted putting said model(s) into MARK for quite some time after the paper was because of fear that people would assume that they could get everything from near nothing ('wow, all I need to do is an occupancy study and I can get an estimate of abundance too! No more need to mark-recapture anything! Wow again!'). As Jim points out, under some pretty (verging on heroic assumptions), you can back out an estimate of abundance. But, you'd need to have a pretty decent way of testing those assumptions (and once you go down that road, you'll see there might be some circularity in the process -- assumptions tested by data that if you have said data, you don't need this approach in the first place).

Remember: just because you can -- in theory -- do something, doesn't mean you should -- in practice.

Or, to paraphrase Gary: 'sharp knives cut stupid people'.

[darryl]

Sometimes they cut smart people too...

[mpwilson09]

Hello,

I have point-count data and to reduce assumptions being made, would you recommend using the multi-state occupancy model instead of the Royle model? Can you get abundance using multi-state though?

Also, I can't seem to find the instructions on the multi-state example that comes in the data package. The excel file is named multistate and it has two columns with numbers ranging from 0 to 2.

Thanks,
Meredith