www.phidot.org

I am working with a dataset that I manipulated into seasons and pooled into weekly surveys. What I am really looking for is a detection probability to use in a different analysis, but the models I have tried in RMark all have high c hat values, from 9 on up. So I changed it back from binary 1/0 indicators for captures to count data and have tried to do a Poisson, Zero inflated Poisson, and Negative Binomial in unmarked to compare AICs. The NB fit best but still gave an alpha value of -0.9.

The variance is very high, even when I just run constants. Although there are lots of 0's and NA's in my data, I don't think ZINB is appropriate for my data set.( I also don't think QAIC can save it as it is so high, but I also don't understand why RMark isn't giving QAIC values even though chat is greater than 1) I feel like I have run out of options and I am in a time crunch. Any advice on how to go about getting a decent detection probability would be appreciated.

Thanks

Sorry, but no amount of 'statistical fussing' can entirely compensate for poor/sparse data. Running the gamut of different model/data types simply to find one that gives you something you can use, without any real motivation for trying one or another, is data dredging.

Worth keeping the following in mind:

'These methods are statistical, not magical' - Darryl MacKenzie

'The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data.' - John Tukey