Models4PT: Fork of DAGitty

This repository is a fork of the DAGitty project, which provides a collection of algorithms, a GUI frontend, and an R package for analyzing graphical causal models (DAGs). While preserving DAGitty’s original functionality and scope, this fork is being customized to support physical therapy (PT) and healthcare research as part of the Clinical Inquiry Network ecosystem.

---

About Models4PT

Models4PT is a dynamic tool for creating and analyzing graphical causal models tailored to the needs of physical therapy clinicians and researchers. It integrates into the Clinical Inquiry Network, which seeks to enhance clinical reasoning, causal modeling, and evidence-based practice in physical therapy. This initiative includes:

1. Stats4PT:
   A statistical education and guidance platform for physical therapy.
   Learn more here.

2. The Clinical Inquiry Fellowship:
   A program for advancing clinical reasoning and critical inquiry in PT practice.
   Learn more here.

3. Collaborative Tools:
   A suite of resources to model complex patient cases, integrate research findings, and explore clinical reasoning frameworks.

---

Enhancements in Progress

Transition to a PT-Focused Tool

• Adapting the platform to address physical therapy-specific use cases.
• Adding customizable templates and libraries for common PT causal models.

Integration of Bhaskar’s Critical Realist Framework

• Explicit modeling of the empirical, actual, and real domains in causal modeling.
• Support for latent variables and deeper structural analysis.

Expanded Inference Methodologies

To support comprehensive clinical reasoning, Models4PT will integrate options for probabilistic causal inference models, including:

1. Induction (Statistical Inference):
   • Analyzing patterns and generalizations from data to support clinical decisions.
2. Deduction (Probability Logic):
   • Exploring deterministic and probabilistic relationships between variables.
3. Abduction (Bayesian Inference):
   • Generating hypotheses and evaluating their likelihood using Bayesian methods.

Improved User Experience

• Enhanced saving/loading functionality using local storage and, eventually, backend storage for collaborative workflows.
• Streamlined interfaces for creating, modifying, and exporting models.

Planned Extensions

1. Hosting the platform at models4pt.com to provide a publicly accessible, PT-centric tool.
2. Adding features to:
   • Identify and visualize confounders relevant to PT interventions and research.
   • Automate suggestions for evidence-based adjustments to causal models.
3. Integration into broader clinical reasoning education and consulting services.

---

Original DAGitty Project

The original DAGitty project is maintained by Johannes Textor and collaborators. This fork builds on their robust foundation to address domain-specific needs in physical therapy.

For more information on DAGitty, visit:

• Website: dagitty.net
• Publications:
  1. Textor, J., et al. (2017). Robust causal inference using directed acyclic graphs. International Journal of Epidemiology.
  2. Ankan, A., et al. (2021). Testing Graphical Causal Models Using the R Package “dagitty”. Current Protocols.

---

Running the GUI Locally

To test or use the current version of Models4PT locally:

1. Clone the repository.
2. Open gui/dags.html in a modern web browser.
3. Follow the instructions in the GUI for saving/loading models.

---

Acknowledgments

We extend our gratitude to Johannes Textor and the DAGitty team for their pioneering work in causal modeling and for making this project possible through open-source collaboration.

---

Contact

For questions about this fork or collaboration opportunities, please contact:

• Sean Collins, PT, ScD
  Peripatetic PT
  Stats4PT
  Clinical Inquiry Fellowship

  Updated as of January 13, 2025