POPAN Standard Errors

[aduarte]

I am a new to mark-recapture analyses. I currently have 4 years of mark-recapture data on a species of toad and my goal is to estimate abundance of the males. The problem is the males in this species only come out 2-5 days after it rains during the breeding season and individuals may only come out once per season. Though they are geographically a closed population, individuals are not always above ground and available for capture so I thought it was more appropriate to treat it as an open population. Also, the data is pretty ragged because of their breeding patterns. In order to “clean up” the data I lumped the data into 1 week occasions (giving me a minimum of 5 occasions per season). I analyzed the data one year at a time using the POPAN analysis using the models p(t)phi(t)pent(t), p(.)phi(t)pent(t), and p(.)phi(.)pent(t). This allowed me to derive a gross abundance estimate each year as well as an abundance estimate per occasion. The abundance estimates per occasion each year seem to be reasonable, however, 2 of the 4 years have gross abundance estimates with ridiculous standard errors (569.6 (SE=35914.9), 77.1 (SE=72372.9)). So my questions are:

1) In the end I’d like to use the abundance estimates per occasion but are the estimates valid considering the standard errors on the gross abundance estimates?
2) Is lumping the data into 1 week intervals a valid process?
3) Why would I get such large standard errors in my gross abundance estimates?

I would greatly appreciate any help.

Adam

[bacollier]

I am a new to mark-recapture analyses. I currently have 4 years of mark-recapture data on a species of toad and my goal is to estimate abundance of the males. The problem is the males in this species only come out 2-5 days after it rains during the breeding season and individuals may only come out once per season. Though they are geographically a closed population, individuals are not always above ground and available for capture so I thought it was more appropriate to treat it as an open population. Also, the data is pretty ragged because of their breeding patterns. In order to “clean up” the data I lumped the data into 1 week occasions (giving me a minimum of 5 occasions per season). I analyzed the data one year at a time using the POPAN analysis using the models p(t)phi(t)pent(t), p(.)phi(t)pent(t), and p(.)phi(.)pent(t). This allowed me to derive a gross abundance estimate each year as well as an abundance estimate per occasion. The abundance estimates per occasion each year seem to be reasonable, however, 2 of the 4 years have gross abundance estimates with ridiculous standard errors (569.6 (SE=35914.9), 77.1 (SE=72372.9)). So my questions are:

1) In the end I’d like to use the abundance estimates per occasion but are the estimates valid considering the standard errors on the gross abundance estimates?
2) Is lumping the data into 1 week intervals a valid process?
3) Why would I get such large standard errors in my gross abundance estimates?

I would greatly appreciate any help.

Adam

Adam,
Not a POPAN expert of a Houston toad expert, but here are a few thoughts on your post to get you thinking. Assuming you have read all the MARK book as you seem to have gotten some simple models to run (what does a p(.), pent(.), phi(.) model work?)

Not sure on your logic for treating it as an open population? Do you mean within a season (probably not) or between years (probably so). I assume between years is what you mean but what you state above regarding closure is related to availability not population openness?

1) Your abundance estimates are likely not accurate, especially as you seem to be showing a order of magnitude change between a couple of years which is surprising.

2) Can not advise you much as we don't know the structure of your data, how many unique individuals, how many captures/recaptures during each year, or each sampling occasion.

3) Guessing here, but you probably have such as sparse dataset in some cases(few recaptures in sampling occasions) that the model you have posited cannot converge adequately so you get a number but it may not be accurate at all. Again, we would need more info.

Why not do some simple, single year abundance models show for each year? If abundance is your actual interest, and you can assume closure within a year, why not first run a set of year-specific closed capture models (Huggins or the like) on the year-specific data first and see what you come up with before trying to incorporate additional parameters associated with open populations and more complex models? Seems like that might be a good starting point for someone who is new to CMR.

Bret