adminR: R in Swiss Official Statistics

• The Bernerhof is a government building, all participants must register in advance, providing their full name and the name of their organisation. Please use the registration form to sign up for the event: adminR Frühling 2026: Registrierung Bernerhof
  Your data will not be passed on to third parties and will be destroyed after the event.
• Please bring a valid ID (passport/ID). Entry cannot be granted without a valid ID.

As usual there will be 🍻, 🍿 and great discussions after the keynote.

Moving a Legacy Code Base to Posit Workbench – Experiences from Government Finance Statistics
Berger Samuel, Federal Finance Administration

Are you working with a growing number of long scripts stored on a network share? Has your production code grown over years and is increasingly hard to maintain? Are you a social scientist with no formal training in software engineering and wondering whether you’re doing it right? Would you like to refactor your code but can’t because ongoing production depends on it? Do you struggle to coordinate multiple team members using R installations on personal laptops?

These issues are familiar to us and motivated the introduction of Posit Workbench – a centralised development and execution environment for R and Python. Over the past three years, the migration to Workbench has been central in an ongoing project to modernise our code. In parallel, we are changing the way we collaborate using version control techniques and promoting best practices for clean and maintainable code among team members. Insights into what worked, what did not (yet) and what we wish we had known earlier.

KOMA – An R package for Bayesian estimation of simultaneous equation models
Scherer Merlin, KOF Institut

Central banks and forecasting institutions rely on simultaneous equation models for their interpretability and coherent, theory-driven forecasts. KOMA (KOF Macroeconomic Analysis) is an R package for Bayesian estimation of such models using Metropolis-within-Gibbs sampling. It provides text-based model specification, customizable prior distributions, full posterior inference, conditional forecasting, out-of-sample evaluation, and ready-to-use tables and plots.