README

This repository accompanies the manuscript Pavlovian Bias is Associated with Symptom Severity but Not Diagnostic Status in Individuals with both Anxious and Non-Anxious Depression (Goldman et al., 2025). It provides the full set of scripts, data, and outputs used to model, fit, and analyze behavioral data from a Go/No-Go task using Reinforcement Learning Drift Diffusion Models (RLDDMs). The repository integrates MATLAB and Python code for computational modeling, along with R scripts and datasets for statistical analysis and visualization.

Data Analysis Files

The data_and_analysis_scripts/ folder contains the following:

• GNG_analyses_5-13-25.Rmd: R Markdown notebook for conducting the main behavioral and statistical analyses reported in the paper.
• corrplotplus.R: Custom R function for generating enhanced correlation matrices with annotated Bayes Factors or p-values, significance stars, and optional clustering.
• GNG_experiment_data_5-6-25.csv: Dataset including computational and descriptive indices of task performance as well as psychological measures.
• GNG_experiment_data_5-6-25_dictionary.xlsx: Data dictionary accompanying the CSV file, detailing column definitions, coding schemes, and variable descriptions.
• param_recov_full_param_distributions.csv: Parameter recovery results using the full reasonable distributions of computational parameter values.
• param_recov_full_param_distributions_dictionary.xlsx: Dictionary file providing explanations of each column in the parameter recovery CSV.

---

Repository Files

MATLAB Scripts

1. GNG_main_hierarchichal_RLDDM.m

   • Main wrapper script for fitting and simulating RLDDMs.
   • Supports hierarchical and non-hierarchical parameter fitting using parametric empirical Bayes.
   • Configurable parameters include learning rates, decision thresholds, drift rates, and biases.
   • Outputs fit results and optionally simulates and re-fits on simulated data.

2. fit_gonogo_laplace.m

   • Function for fitting RLDDMs to Go/No-Go task data.
   • Uses variational Bayesian methods to estimate posterior distributions of model parameters.
   • Handles parameter transformations and hierarchical modeling.

3. GNG_dcm_peb_fit.m

   • Performs group-level hierarchical model fitting using parametric empirical Bayes.
   • Incorporates second-level constraints to refine parameter estimates across subjects.

4. GNG_spm_dcm_fit.m

   • Wrapper for Bayesian inversion of subject-specific DCMs.
   • Supports parallel computation and includes diagnostic output for model performance.

5. GNG_spm_dcm_peb.m

   • Core function for hierarchical inversion of DCMs with second-level design matrices.
   • Implements Bayesian model reduction and variational Laplace for parameter estimation.

6. wfpt.m

   • Computes the probability density function of first-passage times for a Wiener diffusion process.
   • Used for calculating response probabilities in DDMs.

---

Python Scripts

1. run_RL_DDM_test_multiple_models.py
   • Automates testing of multiple RLDDM configurations.
   • Submits Slurm jobs for parallelized fitting across subjects and models.
   • Supports dynamic creation of result directories and custom parameter mappings.

---

Slurm Batch Script

1. run_RL_DDM_test_multiple_models.ssub
   • Template for Slurm job submission.
   • Executes RLDDM fitting for a given subject and model configuration.

---

Usage

1. MATLAB Workflow:

   • Edit GNG_main_hierarchichal_RLDDM.m to configure paths, subjects, and model parameters.
   • Run the script for fitting and simulation.

2. Python Workflow:

   • Edit run_RL_DDM_test_multiple_models.py to specify subjects, models, and result directories.
   • Submit Slurm jobs using run_RL_DDM_test_multiple_models.ssub.

---

Dependencies

• MATLAB: Requires the SPM12 toolbox, including the DEM submodule.
• Python 3.x: Requires standard libraries such as os and subprocess.
• Slurm: Needed for batch processing on high-performance computing clusters.
• R (version ≥ 4.0): Required packages include tidyverse, Hmisc, corrplot, BayesFactor, and glue.
• RStudio (optional): Recommended for running and knitting .Rmd files.
• Microsoft Excel or compatible software: Required to view .xlsx data dictionary and parameter recovery documentation files.

---