In light of recent discussions about the reluctance of authors of published papers to admit their mistakes, I thought I would post four correction notices that I felt necessary to publish.
Since 1999Correct the paper that appeared in 1993:
at 2013Correct the paper that appeared in 2008:
Since 2017Correct the paper that appeared in 2006:
Finally, there was This note was written in 2014. We revise a 1996 paper published in a volume rather than a journal. There is no way to formally notify a correction of a book that has already been published, so I posted the correction on my blog. It’s an interesting story. Simply put, we told the newspaper that we had done something we had never done before and made false statements. In the correction notice, I did a simple simulation in R to confirm that we had indeed made a mistake.
argument
I like how the revision notifications are very clear and unhedged. We were wrong. That’s the story!
My only regret is that I did not investigate how the mistake occurred regarding these notices. If I could do this entirely, I would enable errors by adding a statement to each about what aspect of the workflow failed.
We add the following to the correction notice posted above:
1999: It is not generally clear to us how to avoid false proofs of this kind. The problem is that false statements seem so natural to us that we haven’t thought to look at them carefully.
2013: I think this error could have been avoided if I had used a better data analysis workflow that created correlation graphs between all pairs of variables and a computational workflow that made it easier to fix the problem.
2017: If the paper had included fully working examples along with the code, you might have discovered this issue.
It’s not perfect, but you get the idea.
