Sunday, April 17, 2016

The Crisis in Social Science Continues

This roundup of recent happenings in the ongoing crisis of replication, bad statistics and outright fraud in social science is full of sad news: more famous results retracted, whole classes of results invalidated by poor statistical controls. The debate over Australia's experiment with gun control continues, with a new study arguing that the earlier studies that concluded the reform did reduce violence were wrong and there was no effect. One recent article which looked at why experimental results can't be replicated concludes that
the influence of questionable research practices is at the heart of failures to replicate psychological findings, especially in social psychology.
But the worst part is an account of social scientists who found serious errors in their own published work and tried to retract it. One journal told an author that they charge $10,000 to publish retractions. Another was told, by the American Sociological Review, that they had no mechanism for retractions, but the authors were welcome to submit their new findings as a separate paper. They did so, but the new paper was rejected with the usual form letter saying the ASR "publishes only the very best submissions." Which was grimly amusing to authors who have realized that the paper the journal did publish contained an invalidating error.

It's disheartening.

1 comment:

G. Verloren said...

"The debate over Australia's experiment with gun control continues, with a new study arguing that the earlier studies that concluded the reform did reduce violence were wrong and there was no effect."

How exactly could this possibly be improperly measured? Either there were less incidents of gun related violence after the changes to gun control, or there weren't.