delphyr

Reproducible Delphi analysis, independent of the application.

delphyr validates protocols and response scales, analyses frozen round data, and provides the transactional services used by delphyrApp. Offline analysis runs without Shiny or a database. Study services use PostgreSQL.

This is development software for synthetic studies. It is not a CRAN release, a production qualification, or a validation of a study's scientific choices.

Install and start

# install.packages("remotes")
remotes::install_github("CTTIR/delphyR", subdir = "packages/delphyr",
                        build_vignettes = TRUE)

library(delphyr)
analysis <- analyse_round(demo_snapshot())
analysis$decisions
analysis$denominators
analysis$missingness

The synthetic snapshot explicitly uses pooled consensus. The complete demo protocol instead requires adequate results in all designated groups. Example thresholds are not recommendations. Missing responses are not zero ratings.

Follow the worked example

vignette("offline", package = "delphyr")

The offline vignette executes two-round analysis, contrasts group rules, reports missingness, checks comparability, and prepares suppressed feedback. A feedback draft still needs a reviewed release before participant use.

reproduce_export(path) verifies a previously obtained private round export and recomputes its analysis without the application or Quarto; reproduce_study_export(path) does the same for every round of a study-level research export and for the comparisons between rounds. Exports do not include qualitative originals or account mappings; they are not automatically suitable for public release.

Study services

The repository guide covers the local PostgreSQL environment, worker, application, and verification commands. Services recheck identity, study membership, capability, and state. Save receipts follow database commits; submitted revisions and released scientific records remain traceable.

Panel imports create unbound drafts. Synthetic invitation acceptance requires explicit approval of an existing issuer/subject identity. Campaigns produce only local sink receipts. These boundaries do not qualify real recruitment, external messages, or production identity infrastructure.

See implementation status for executed checks and remaining acceptance gates. Licensed under the MIT License.

Use of LLM tools

Where applicable, LLM tools were used for narrowly defined, non-authorial tasks: copyediting, prose refinement, Markdown/LaTeX formatting, boilerplate scaffolding (such as CI configurations and build scripts), and code refactoring. Tools included Chat AI, the KISSKI (GWDG) service, and self-hosted Mistral Small (24B, Apache-2.0) served through a local vLLM architecture on an Ubuntu Linux workstation with Intel processors and NVIDIA GPUs. The self-hosted model ran locally, with no data sent to third parties.

All scientific claims, methodological choices, analyses, interpretations and conclusions remain the authors' own. Where applicable, LLM-generated material must be reviewed and revised before inclusion, and references must be verified against their DOI, arXiv ID or ISBN.

Local software setup: vLLM with Python/PyTorch, plus R and RStudio.