openCR - Parameter values constant across sessions in model output
Emily Seccombe
Hello,
Please could you help with the following query re an openCR model output?
In my model output, I am seeing the output of parameters (lambda0, phi, f and sigma) are the same in each session. I had expected the values to change between sessions. Is my expectation incorrect, or have I set up the model incorrectly?
Additionally, I have the warning message: In openCR.fit(capture_hist_object_correct_order, type = "JSSAsecrf", : possible maximization error: nlm returned code 3. See ?nlm
I understand that this is not always a problem, but am struggling to know how to determine if this indicates an issue in my case.
Details of my data and model are below. Attached is a file with some output from the model. I can supply details of the code directly. Please let me know if more information is required to help troubleshoot this.
Many thanks in advance for any help!
All the best,
Emily
----------------
Data is from repeated captures of slow worms at a site. Traps are ACOs (reptile survey sheets).
Session is a combination of year and season (e.g. Spring 2011). In most years, trapping occurred in both Spring and Autumn.
This results in the following sessions:
2011spring, 2011autumn , 2012spring , 2015spring, 2015autumn, 2016spring, 2016autumn, 2018spring, 2018autumn, 2019spring, 2019autumn, 2021spring, 2021autumn
Each session is closed (we can assume no change in population).
Each session has a variable number of secondary sessions (occasions) within it when recording occurred.
The exact location of traps changes between sessions.
An openCR fit model was run using the JSSAsecrf type.
model_JSSAsecrf_hhn <- openCR.fit(capture_hist_object_correct_order,
type = "JSSAsecrf",
mask = mask,
detectfn = "HHN",
trace = TRUE,
ncores = 32
)
JSSAsecrf was selected because: we want JSSA over CJS, as JSSA models do model the first capture of each animal, and lead to estimates of abundance and recruitment; we want secr because we want to do a spatial model; we want recruitment parameterisation 'f', because we want the default entry probability, rather than any other recruitment parameter; and we want full likelihood not conditional likelihood so we do not add 'CL'
Detectfn used was HHN. It is intended to model other detection functions and compare AICs subsequently.
Here is a summary of the traps and capthist:
With this model, the model output from print(model_JSSAsecrf_hhn) gives the same values of the fitted parameters for every session:
e.g. for lambda0:
The values of phi, f and sigma are also constant over each session.
I had expected the values to change over each session. Is my expectation incorrect, or have I set up the model incorrectly?
