Four Best Practices for Using ChatGPT/Claude as an Effective R Assistant
Bulletin de la Dialyse à Domicile · 15.06.2026 · 9 (2) · 103-114
DOI: 10.25796/bdd.v9i2.87112 ↗Kurz gefasst
Fachartikel in der Zeitschrift Bulletin de la Dialyse à Domicile, online erschienen am 15.06.2026. Verfasst von Claire Della Vedova. Die Arbeit nennt 6 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Artificial intelligence can be of considerable help to those wishing to publish their work; however, using it without a critical eye can lead to completely false results. We felt it would be useful to ask a biostatistician and trainer to raise our readers’ awareness of the potential pitfalls of such software when analysing their data, using the example of a programme widely used in medical research.This article presents four best practices for the effective use of language models (LLMs) such as ChatGPT or Claude in an R programming context. The author begins by describing the three most common errors produced by LLMs: the invention of non-existent functions, the use of obsolete syntax and, even more seriously, the suggestion of inappropriate statistical methods. These errors can go unnoticed by beginners and lead to erroneous analyses. It is therefore essential to systematically check the suggested functions, pay attention to R’s warning messages, and understand the conditions under which the recommended statistical tests apply.The article then details a three-step method for obtaining more reliable results: providing a precise context regarding the data and the working environment, clearly describing the desired objective, and specifying the expected format of the output. The author places particular emphasis on data confidentiality.A third section explains how to verify that generated code actually works. Four checks are suggested: checking for the absence of errors and “warnings”, verifying the structure of the results, assessing the plausibility of the values obtained, and testing the code on a small set of known data.Finally, the article specifies when to prioritise an LLM, when to consult the official R documentation, and when to seek the advice of a competent human. The overall aim is to empower the user whilst using LLMs as support tools rather than as substitutes for critical thinking.
Zitieren
APA
Vedova, C. D. (2026). Four Best Practices for Using ChatGPT/Claude as an Effective R Assistant. Bulletin de la Dialyse à Domicile, 9(2), 103-114. https://doi.org/10.25796/bdd.v9i2.87112
Vancouver
Vedova CD. Four Best Practices for Using ChatGPT/Claude as an Effective R Assistant. Bulletin de la Dialyse à Domicile. 2026;9(2):103-114. doi:10.25796/bdd.v9i2.87112
BibTeX
@article{Vedova2026,
author = {Vedova, Claire Della},
title = {Four Best Practices for Using ChatGPT/Claude as an Effective R Assistant},
journal = {Bulletin de la Dialyse à Domicile},
year = {2026},
volume = {9},
number = {2},
pages = {103--114},
doi = {10.25796/bdd.v9i2.87112}
}
EUVIMED