ChatGPT as a Tool for Biostatisticians: A Tutorial on Applications, Opportunities, and Limitations

Area of Science:

• Biostatistics
• Artificial Intelligence
• Computational Statistics

Background:

• Large language models (LLMs) are transforming various professional fields, including biostatistics.
• LLMs provide assistance in tasks like study plan drafting, code generation, and report writing.
• Potential LLM "hallucinations" pose a risk of introducing inaccuracies into statistical work.

Purpose of the Study:

• To illustrate the impact of LLM applications on contemporary biostatistical tasks.
• To explore the risks and opportunities associated with AI in biostatistics.
• To provide guidance on the responsible use of LLMs in the field.

Main Methods:

• Tutorial-based exploration of LLM applications in biostatistics.
• Analysis of potential inaccuracies and their impact on statistical integrity.
• Discussion of strategies for mitigating risks and leveraging opportunities.

Main Results:

• LLMs can enhance efficiency in biostatistical workflows.
• Inaccuracies from LLM hallucinations can lead to erroneous statistical statements and conclusions.
• Careful verification of LLM outputs is crucial for maintaining precision and transparency.

Conclusions:

• Advanced LLM applications in biostatistics require sufficient user background knowledge.
• Consistent verification of LLM-generated content is necessary for building calibrated trust.
• Responsible integration of AI tools is key to advancing biostatistical practice.