Quick start
===========

Installation
------------

pyLBA requires Python 3.10 or newer. Install from the repository:

.. code-block:: bash

   pip install git+https://github.com/rgs2151/pyLBA.git

Simulation
----------

Create a model and specify the starting-point bound, thresholds, drift
parameters, and non-decision time:

.. code-block:: python

   from pyLBA import LBAModel, LBAParameters

   model = LBAModel()
   parameters = LBAParameters(
       A=1.0,
       b=[2.0, 3.0],
       v=2.0,
       s=1.0,
       tau=0.1,
   )
   data = model.generate_data(500, parameters, seed=42)
   print(data.head())

A scalar is shared across accumulators. A vector supplies one value per
accumulator. The number of accumulators is inferred from parameter vectors;
pass ``n_acc`` explicitly when all parameters are scalar.

The result is a pandas DataFrame with response times in ``rt`` and zero-based
choice labels in ``response``. A fixed seed reproduces the simulated data.

Fitting and prediction
----------------------

This example estimates thresholds while holding the other parameters fixed:

.. code-block:: python

   model.fit_mcmc(
       data,
       priors={"A": 1.0, "v": 2.0, "s": 1.0, "tau": 0.1},
       draws=1000,
       tune=1000,
       chains=4,
       random_seed=42,
   )
   predictions = model.predict(100, seed=7)

   import arviz as az
   print(az.summary(model.trace, var_names=["b"]))
   print(model.parameters)

``trace`` contains the posterior samples. ``parameters`` contains posterior
means, and ``predict`` simulates at those means. See :doc:`fitting` for default
priors and custom distributions.

Plot simulated data
-------------------

.. code-block:: python

   import matplotlib.pyplot as plt

   for response, trials in data.groupby("response"):
       plt.hist(trials.rt, bins=30, alpha=0.5, label=f"Response {response}")
   plt.xlabel("Response time")
   plt.ylabel("Trials")
   plt.legend()
   plt.show()