Simulating capture histories, and estimating paramters, under misidentification

Code to recreate analysis "Evaluating tradeoffs between automation and bias in population assessments relying on photo-identification," which is in press at Conservation Biology.

Simulation

The simulation proceeds in three general steps.

1. Simulate capture histories using a POPAN model.
2. Corrupt capture histories with misidentification error, using predetermined rates.
3. Estimate demographic parameters by fitting capture-recapture models to the corrupted histories.

Configuration

Users should have some form of conda installed (I prefer miniforge). Then, they need to set up a conda environment using the requirements.txt file. For example,

conda create --name <env> --file <this file>

The modules rely on config files, in the form of .yaml. Users can provide their own configs (TODO: Improve documentation for this process). Otherwise, they can recreate the results from the paper using the scr.config script, which creates .yaml files in the the config directory. The scr.config script is called from the command line, for example,

python -m src.config 

Simulating capture histories

The src.simulate script simulates data under an open population capture recapture model (by default, a POPAN model, although the CJS model is available too). Then, it corrupts the capture histories using the misidentification process. Users can run the script from the command line, specifying the strategy with the --strategy argument,

python -m src.simulate --strategy check_0
python -m src.simulate --strategy check_5
...
python -m src.simulate --strategy check_25

The script uses the POPAN class, from the model module, and its simulate() function to simulate a capture history. Then it corrupts the history with misidentifications, using the MissID class from the miss_id module. By default, simulate simulates 100 replicates for dataset in a given strategy.

Estimating demographic parameters

The src.estimate script accomplishes task #3 above for a given scenario and model. It relies on the POPAN class, from the model module, and its estimate() function. Users can run the script from the command line, specifying the strategy with the --strategy argument,

python -m src.estimate --strategy check_0
python -m src.estimate --strategy check_5
...
python -m src.estimate --strategy check_25

The parameters are estimated using PyMC.

Collating the results

Finally, users can collate the results into a .csv file, which contains relevant statistics from the posterior distribution of each parameter. To do so, Users can run the script from the command line, specifying the strategy with the --strategy argument,

python -m src.analyze --strategy check_0