Skip to content

Home

Jump to bottom
Samuel Jenness edited this page Oct 12, 2026 · 43 revisions

Welcome to the EpiModel Research Lab wiki. It is written for lab members (graduate students, postdocs, and staff) who build and run models with EpiModel and our extension packages, and it may help others who use the same tools. It covers how we develop models on our own computers and how we run them at scale on the Emory RSPH high-performance computing (HPC) cluster.

If you are new, work through the sections in order. Each page starts with who it is for and what to read first. Terms you do not know are in the Glossary.

0. Start here

Onboarding checklist

Some of these take several days, so start them on your first day.

  • Emory NetID and Emory VPN access.
  • A GitHub account, added to the EpiModel organization. Ask Sam. You need this to see our private repositories, including EpiModelHIV-p.
  • Access to the ARTnet data (see ARTnetData).
  • An RSPH HPC account (see HPC setup).
  • An invitation to the lab Slack workspace.
  • Software installed on your computer (see Setting Up Your Computer).

Getting help

  • Questions about our code and workflow (EpiModel, EpiModelHIV, renv, Spack, Slurm, slurmworkflow): ask in #coding-help on the lab Slack.
  • Bugs in one of our packages: open an issue on that package's GitHub repository.
  • HPC account, password, VPN, or cluster problems that are not about our code: email RSPH IT at help@sph.emory.edu, starting the subject line with "RSPH HPC cluster:" (see HPC Rules, Storage, and Getting Help).
  • Something on this wiki is wrong or unclear: fix it or tell us (see Contributing to This Wiki).

1. Set up your computer

We write and test all code locally and use the HPC only to run it at scale, so your own computer comes first.

2. Learn EpiModel

If you are new to network modeling, start here before your first project.

TERGM

Our network models are temporal exponential-family random graph models (TERGMs), fit with netest() and checked with netdx(). For background on ERGMs, TERGMs, and model formulas, see the course chapters on ERGMs, TERGMs, estimation, and parameterization from egocentric data.

3. Start your first applied project

  1. Applied Modeling Project Structure: the project repository, your EpiModelHIV-p branch, and what happens on your computer, on GitHub, and on the HPC.
  2. Getting Started with EpiModelHIV: create your project from the Template, run a first model, create your EpiModelHIV-p branch, and open your pull request.
  3. The EpiModelHIV-Template README: the step-by-step setup and the four project chapters (networks, model development, calibration, interventions).
  4. Using renv for Reproducible R Environments: how each project keeps its own package versions, on your computer and on the HPC.

ARTnetData

Our HIV models are parameterized from the ARTnet study. The data are in the private ARTnetData package. To get access, email Sam to request the ARTnet data use agreement; once it is signed, you are given access on GitHub. Details are in the ARTnet README. Without access, installing a project's packages fails with an error about ARTnetData.

EpiModelCOVID projects use the same two-repository pattern, with EpiModelCOVID as the software repository. There is no separate tutorial; ask Sam for a current example project.

4. Develop model code

5. Use the HPC

HPC setup

Read these in order. Never run R or any computation on the login node.

  1. Registering for and Logging in to the HPC
  2. Setting Up Your HPC Environment: SSH keys, the ssh sph shortcut, and aliases.
  3. HPC Rules, Storage, and Getting Help: cluster policies, where to put files, quotas, and backups.
  4. Working with Files on the HPC
  5. Using Git on the HPC
  6. Using R via Spack on the HPC

Slurm basics

All work on the cluster runs through the Slurm scheduler. Our projects generate their jobs with slurmworkflow (next section), but every HPC user should first learn to run R through plain Slurm: interactive sessions with srun, batch jobs with sbatch, including passing parameters to a script, and checking and canceling jobs with squeue and scancel.

  1. Running R Interactively with srun: partitions, resource requests, and an interactive session.
  2. Running R Batch Jobs with sbatch: job scripts, job arrays, and passing parameters with --export. The sbatchExample folder in the EpiModeling repository has the complete example.
  3. Monitoring and Managing Slurm Jobs

slurmworkflow

Project jobs are generated on your computer with slurmworkflow and EpiModelHPC, copied to the cluster, and started there.

  1. Running Project Workflows on the HPC: build, deploy, monitor, and download a workflow.
  2. Running Network Models with Checkpoints

Background: the slurmworkflow documentation and the EpiModelHPC article Using slurmworkflow with EpiModel.

6. Calibrate and analyze

  1. Calibrating a Model: targets, ballpark runs, and automated calibration with swfcalib.
  2. Making a Restart Point: the saved burn-in that calibration and intervention runs start from, and when to make a new one.
  3. Programmatically Defining Scenarios
  4. Processing Intervention Results: outcomes, infections averted, and formatted tables.

The Template's calibration guide lists every file in chapter C, and the swfcalib documentation explains the package.

7. Publish

8. Admin notes

For the people who maintain the lab's shared HPC setup and this wiki. You do not need these to do your research.

EpiModel Research Lab wiki. Something out of date or unclear? Post in #coding-help on the lab Slack, or fix it yourself: see Contributing to This Wiki.

Home

0. Start here

1. Your computer

2. Learn EpiModel

3. First project

4. Model code

5. HPC

6. Calibrate and analyze

7. Publish

8. Admin

Clone this wiki locally