Repeating probability of authors with retracted scientific publications

ABSTRACT

Both the scientific community and the general public have expressed concern over scientific misconduct. The number of retracted articles has increased dramatically over the past 20 years and now comprises about .02% of the 2 million articles published each year. Retraction of publications available in large public databases can be analyzed as an objective measure for scientific misconduct and errors. In this project, we analyzed retractions of scientific publications using the Web of Science (WoS) and PubMed databases. We found that a power law is applicable to distributions of retracting authors and retracted publications with exponents of about −.6 and −3.0, respectively. Application of a power-law model for retracted publications implies that retraction is not a random event. Analysis of the retraction distributions suggests that a small fraction (1–2%) of retracting authors with ≧5 retractions are responsible for around 10% of retraction. The probabilities for their repeating retraction are calculated using a statistical model: 3–5% likelihood of repeat retraction for authors with a single retraction at five years after the latest retraction and 26–37% for authors with five retractions at five years after the latest retraction. By focusing on those with repeated retractions, this analysis could contribute to identification of measures to reduce such repetition of retractions.

KEYWORDS:

• Power law
• repeating probability
• research misconduct
• retraction of publications

Acknowledgment

The authors appreciate Dr. Zoë Hammatt for her editing and critical reading.

Disclosure statement

The authors have no conflict of interest related to any aspect of this work.

Supplemental data

Supplemental data for this article can be accessed here.

Additional information

Funding

This work has no financial support.