← Back to research

Podium · Western Institute of Nursing · 2025

To chat or not to chat: A comparative analysis of ChatGPT and a statistical program

Tolentino, D. A., Boy, P., Long, S. N., Bilog, A. D., Kohout, E., Skrine Jeffers, K., & Shabaik, H. (2025). Published in Communicating Nursing Research, 58.

Artificial intelligenceResearch methodsData analysis
The big idea

Can an AI chatbot analyze research data as reliably as standard statistical software? We compared the two head to head.

Who
A synthetic dataset of 1,339 records
How
Compared ChatGPT-4 with SPSS on the same analyses
What we learned
How similar and reliable ChatGPT’s results are

In plain language

Researchers are starting to ask whether AI chatbots can help with data analysis. We compared ChatGPT-4 (the paid version) with SPSS, a standard statistical program, running the same analyses on the same dataset to see how similar and reliable the results were. The goal was to test, rather than assume, whether these tools are trustworthy for nursing research.

About this study

We used a synthetic cross-sectional dataset (n = 1,339) on insurance charges with seven variables (such as age, gender, and BMI) and compared analyses run in ChatGPT-4 against the same analyses in SPSS, assessing the similarity and reliability of the results.

Key themes

1

AI as an analyst

Large language models can now run data analyses, tested head-to-head against established software.

2

Trust but verify

Accuracy and reliability for research remain uncertain.