Cervical Cancer Prevention Agent-Based Model

This model simulates the impact of several interventions on cervical cancer outcomes. The primary outcomes are cervical cancer incidence, mortality rates, and intervention costs. A secondary outcome (for some countries) is HPV prevalence. The interventions of interest include HPV vaccination, cervical cancer screening, and HIV treatment.

The simulation is implemented using python and primarily numpy based state classes.

Environment Setup

There are several ways to work with project code locally.

VirtualEnv Setup

If you already have python and pip installed on your computer, you can run the following from the repo directory:

pip install virtualenv
virtualenv python_env
source python_env/bin/activate
pip install -r docker/requirements.txt
pip install -e .

Docker Setup

docker-compose build

Running Scenarios/Iterations

Terminology

An iteration is a single model run, simulating the lifetimes of one group of women. It is characterized by a random seed and a set of model parameters.
A scenario is a collection of iterations that share the same set of model parameters but differ in their random seed. The purpose of a scenario is to observe how much variation one might expect in the model results given a particular set of parameters.
An experiment is a collection of scenarios whose model parameters differ. The purpose of an experiment is to compare the model results between different sets of model parameters.

Run.py

We will now provide examples for how to run scenarios or iterations, using Zambia as an example.

You can run one iteration of a scenario with:

python run.py experiments/zambia/scenario_base

You can run one x iterations of a scenario and use multiprocessing with:

python run.py experiments/zambia/scenario_base --n=x --cpus=2

Output

Running the experiment will produce several files and directories:

An overall log file will be created in the experiment directory.
One directory per iteration will be created inside each scenario directory. The iteration directory will contain:
- Any output files containing the model results for that iteration.
- A log file for that iteration.

Running the tests

To run the test suite, execute the following command from the project's root directory:

pytest

Profiling

sudo pyinstrument --html -o pyinstrument_test.html experiments/zambia/src/example.py

RTIInternational / cervical-cancer-abm