Usage¶
LASIK ships as both a command-line tool (metapop) and an importable Python
package (laser.cholera). This page covers the minimal "just run the
defaults" path. For deeper workflows — overriding parameters, scoring
against observed data, building configurations from scratch — see the
tutorials, the how-to guides,
or jump directly into the parameter reference.
For installation see Installation. For the full API see the API reference.
Quick start¶
From the command line¶
Once installed, the metapop entry point runs the bundled cholera
simulation:
Common flags (see How-to › Override parameters for the full list):
--params <path>— load parameters from a JSON or.json.gzfile instead of the bundled defaults.--seed <int>— set the PRNG seed (default20240930).--viz/--pdf— display matplotlib visualisations or write a PDF.--outdir <path>— directory for any HDF5 or PDF outputs.--over key:value— override a single parameter, repeatable.--loglevel {DEBUG,INFO,WARNING,ERROR,CRITICAL}— verbosity.-q/--quiet— suppress the per-tick progress bar.
From Python¶
run_model returns the populated
model object:
from laser.cholera.metapop.model import run_model
# Use the bundled defaults
model = run_model(None)
After run_model returns, the per-tick compartment state is on
model.people (e.g., model.people.S) and the per-patch derived quantities
are on model.patches. The final-tick model log-likelihood (when enabled
via the calc_likelihood parameter) is on model.log_likelihood.
Where to look next¶
- Tutorials — first run, single-location, multi-location country.
- How-to guides — override parameters, enable vaccination, configure mobility, enable seasonality, calibrate and score, interoperate with MOSAIC.
- Reference — parameter index, API reference.
- Configurations — single-location, Mozambique adm-2, SSA baseline.
- Explanation — model overview, transmission, mobility, reporting and likelihood.
Cross-checking against the upstream R reference¶
The R implementations the Python port was translated from live under
reference/ at the repo root (not packaged into the sdist). They are useful
for sanity-checking the Python behaviour when something looks off.
One-time setup¶
Install testthat and the upstream MOSAIC R package:
Rscript -e 'install.packages("testthat")'
Rscript -e 'remotes::install_github("InstituteforDiseaseModeling/MOSAIC-pkg")'
If MOSAIC is already on your R library path, you can skip the second
command. Verify with:
Run the R test files¶
The R tests are free-standing testthat files; invoke them directly:
Run from the repo root so the relative path resolves.
Caveats
MOSAICmust be recent enough to include any upstream commits you want to cross-check against (e.g., the in-window peak filter).- R's
set.seed(123)does not produce the same RNG stream as Python'snp.random.default_rng(123). Compare structural properties (finite, ordering, inequality) across the two, not exact numerical values — the Python test files already follow that convention.