README

Note: This repository is currently under development.

Do not use this code in its current state. Features and functionality are incomplete and subject to change.

---

Model Fitting for Theory of Mind (False Belief) Task

This repository contains MATLAB and Python scripts for modeling, fitting, and analyzing False Belief Task (FBT) data using advanced computational models. The scripts support parameter estimation, hierarchical modeling, and batch processing for subject-level and group-level analyses.

---

Repository Files

MATLAB Scripts

1. FBT_run_models.m

   • Main script for model fitting across multiple configurations.
   • Iterates over parameter settings (alpha, beta, delta, lambda) and outputs results in .csv and .mat formats.
   • Includes visualization of model fit performance.

2. FBT_config.m

   • Configuration script for FBT model fitting.
   • Defines parameter bounds, priors, and input data paths.
   • Supports both local and Prolific datasets.

3. FBT_llfun.m

   • Model-specific likelihood function for agent-specific learning in FBT.
   • Calculates posterior probabilities, subject beliefs, and predicted responses.
   • Supports both Maximum Likelihood (ML) and Maximum a Posteriori (MAP) estimation.

4. emfit_CMG.m

   • Expectation-Maximization (EM) fitting script for FBT models.
   • Includes functionality for computing subject- and group-level parameter estimates, Bayesian model comparison, and plotting belief trajectories.

---

Python Scripts

1. runall_FBT.py
   • Automates batch fitting of FBT models using Slurm.
   • Dynamically creates result directories and logs.
   • Configures parallel processing across subjects.

---

Usage

1. MATLAB Workflow:

   • Edit FBT_config.m to specify parameter bounds and data paths.
   • Run FBT_run_models.m to fit models and visualize results.

2. Batch Processing with Python:

   • Modify runall_FBT.py to define subject lists and results directory.
   • Execute to submit Slurm jobs for parallelized fitting.

---

Dependencies

• MATLAB: Required for core model fitting and analysis.
• Python 3.x: Used for job automation and directory management.
• Slurm: Needed for high-performance batch processing.

---