The Journal of Multivariate Analysis has retracted a paper it was never meant to publish — a problem, it seems, of multivariate analyses.
The article, titled “Regression estimation with locally stationary long-memory errors,” came from a pair of statisticians in Chile, Wildredo Palma and Guillermo Ferreira.
It appears that the article did not pass muster with the reviewers, but that the editor somehow missed the message. As the retraction notice explains:
This article has been retracted at the request of the Editor-in-Chief, who noted an administrative error. The EiC accidentally accepted while the reviewers’ recommendation was to reject. We apologize that this was not detected during the submission and review process.
The editor in chief of the JMVA is Jan de Leeuw, a UCLA statistician and founder of the department. We spoke with de Leeuw, who told us that the article had been a “straight rejection” — that is, no possibility of revision for eventual acceptance — at the level of both the reviewers and the associate editor.
My idea was that I would follow that and reject the paper but for some reason it wound up in the editorial system as an accept.
Whether the moment of error occurred when de Leeuw clicked the wrong button on his screen, with the editorial office in India or whether the mixup happened with the online editorial manager itself, he doesn’t know.
Once a paper is approved, he says, an article is out of his hands completely as it wends its way through the Elsevier production process.
de Leeuw said one of the reviewers, who had written a particularly negative review, alerted the associate editor, who informed him that the paper had been published.
de Leeuw moved to retract. The authors were unhappy — “They had every reason in the world not to be happy,” he said — and urged him to reverse the retraction.
They asked us to reconsider the retraction, and we didn’t because it was clear that it was clerical error.