Fixing detectability across all sites

[CJWM]

Greetings,

I have inherited some point count data with which detection probabilities have already been calculated using a removal model. The data is set up so that once a species is encountered the surveys stop. So, if a cardinal is detected in the first occasion, the data reads 1,0,0,0,0 even if the bird was encountered in another occasion. If it is first detected in the second occasion the data reads 0,1,0,0,0 and so on. The detection probabilities are assumed to be constant across all sites. I’d like to collapse the 5 occasions into one occasion and analyze this data in PRESENCE, but I need a way to make detectability a constant number across all sites. Is this a reasonable way to go about analyzing this dataset? And is it even possible?

Thanks,
Chris

[dhewitt]

Was detectability estimated from the same data set that you've inherited? If so, it would be best to estimate det. and occ. at the same time, rather than subbing det. in as fixed. Perhaps you have det. estimates from another study that you're using here? Even in that case, you'll be assuming that the error in det. is zero, which would only be reasonable if it's got a tiny standard error.

[darryl]

Greetings,

I have inherited some point count data with which detection probabilities have already been calculated using a removal model. The data is set up so that once a species is encountered the surveys stop. So, if a cardinal is detected in the first occasion, the data reads 1,0,0,0,0 even if the bird was encountered in another occasion. If it is first detected in the second occasion the data reads 0,1,0,0,0 and so on. The detection probabilities are assumed to be constant across all sites. I’d like to collapse the 5 occasions into one occasion and analyze this data in PRESENCE, but I need a way to make detectability a constant number across all sites. Is this a reasonable way to go about analyzing this dataset? And is it even possible?

Thanks,
Chris

Why do you want to collapse the data? Do you have repeated surveys at each point count station?

[CJWM]

I do have repeated surveys at each site, but the way the sampling was done, I only know the occasion in which the bird was first observed. Whether or not it was observed in the following occasions is unknown. The detection probabilities that were generated from this data used a removal model. PRESENCE doesn’t seem to handle this type of data well because it keeps giving me huge standard errors. So, I figured I could collapse all occasions into one to get an idea of presence or absence at a point, and then apply the detection probability constantly across all sites in order to generate an occupancy model.

Chris

[darryl]

Are the detection probabilities that have been calculated previously at the individual or species level? My guess (though I could be wrong) is that if they've used the removal model eg Farnsworth et al, it will be at the individual bird level. You can't take one type of detection probability and apply it directly the other situation.

The fact that PRESENCE is giving you huge SE's may indicate something more problematic with your model, or how you're setting up your data. Are you getting any warnings? In PRESENCE with removal-type data you should indicate a missing value (ie "-") for surveys after the first detection because you do not know if it was a 0 or a 1.
Darryl

[CJWM]

Thanks for your help, i was under the impression that PRESENCE couldn't analyze removal type data. I simply replaced the false 0's with "-" and it seems to be working well.

Thanks again,

Chris