Computationally Reproducible Research
Methods seminar – learning computational reproducibility by doing computational reproductions
Rachel Heyard 
Aim
In this methods seminar, participants first build the conceptual vocabulary and a theoretical understanding of what computational reproducibility entails. They then apply and refine this understanding by attempting a computational reproduction of a published paper. We will focus on papers presenting the results of meta-analyses and participants will need to use the material presented in the paper to write simply R code to verify that the numbers (or other types of results, e.g., forest plots etc) agree.
The results of all reproductions will be synthesised into a joint manuscript and submitted to Replication Research (R2), a diamond open-access journal. All participants and instructors will be offered authorship. The manuscript will be finalised and submitted after the seminar ends; students who volunteer to take a leading role in the finalisation will be given lead authorship positions.
The seminar design follows Seibold et al. (2021), who ran a similar exercise as a master’s statistics course.
Learning objectives
By the end of the seminar, seminar participants will be able to:
- Name the key properties of computational reproducibility and know the tools that enable computationally reproducible work.
- Write a complete (reproduction) protocol.
- Use Quarto and GitLab as a reproducibility toolkit.
- Recognise basic statistical and data red flags of irreproducibility and untrustworthiness.
- Work in a reproducible way; by learning it the hard way, through practicing and reproducing.
- Know the practical mechanics of submitting a reproduction to a diamond open-access venue like Replication Research (RR).
Practical information
| When | Wednesdays, 15:00–17:00 |
| Dates | 23 Sept · 7 Oct · 14 Oct · 18 Nov · 25 Nov · 2 Dec · 9 Dec 2026 |
| Where | HIT-E-03 |
| Tools | Quarto, GitLab (UZH), R |
How the seminar works
The seminar is built around two activities, both carried out in groups of two participants (students can also work alone if they prefer):
One student-led input per group. Each group gives a short input of about 10 minutes, based on one resources provided by the instructor. The format is an informal journal club: materials are distributed to the input group and to all other students one week in advance, and there is no need to send slides beforehand.
One computational reproduction per group. Each group selects a target paper from a pre-selected pool, writes a reproduction protocol following a provided template, locks the protocol on GitLab before execution, executes the reproduction, and logs every deviation or ambiguity along the way. Additionally, groups are paired for a mid-course peer review of each other’s repository and protocol.