Independent QA of BUGS code

drnickisaac / testSTINGindicators

Testing indicators for the STING project

MIT License

0 stars 0 forks source link

Independent QA of BUGS code #3

Closed drnickisaac closed 1 year ago

drnickisaac commented 1 year ago

Input from someone with expertise in BUGS language to assess whether code is appropriate for the task and faithful to the model in STING1 report.

drnickisaac commented 1 year ago

Issues that occur to me:

[x] Are the hyperpriors on alpha.p redundant? This code was derived from a model in which there was an annual value.
[x] Is beta3 redundant? I think the two alpha.p parameters make it unnecessary. Perhaps more identifiable if JulDate was centred on zero.
[x] muZ is doing two different jobs: one as probability of occupancy (i.e. bounded between 0 and 1) and the intensity, which should be unbounded above zero. Need to include a cloglog link somewhere to achieve this.

drnickisaac commented 1 year ago

Since setting up a model in Nimble, I have completely rewritten the code, meaning that some of the issues above disappear. I started really simple then built it up. At the time of writing I now have an IDM, linked through the phenology of detection. Some issues that need attention/help from others:

[x] I don't have a cloglog link. I think this is only necessary if I have covariates (i.e. sharing info on the state process, not the observation process). Is that right?
[x] I would like to add covariates .....
[ ] I'd also like to add that detection in the occupancy part should be dependent on lambda[i]

drnickisaac commented 1 year ago

Closing this issue since the mainly reflect the early stages of development.