Despite public pledges, leading scientific journals still allow statistical misconduct and refuse to correct it

Originally published at: https://boingboing.net/2019/02/15/pre-specified-outcomes.html

7 Likes

Medical journals still worship at the altar of frequentist statistics with the p-value being the holy grail. Bayesian statistics are being considered now, especially in oncology studies since a myriad of combinations of different drugs must be evaluated.

Also, rejecting the null hypothesis with a statistically significant p-value doesn’t prove the drug you’re testing is the cat’s pajamas. We have “fiction” and “non-fiction” books, not “fiction” and “truth” books. Strictly speaking, in the criminal justice system folks are “innocent” (the null hypothesis) or “not-innocent.”

9 Likes

Some of these disciplines, like psychology, are just rotten to the core. Last time a statistician tried to point out bad practices on the part of a psychologist, the psych community rallied around the offender, and not only refused to accept that there was anything wrong with the findings, but referred to the criticism as " methodological terrorism." We need to stop calling these people "scientists " and start calling them out on thei bullshit.

6 Likes

Well of course - otherwise, the sheeple might find out about the thimerosal in the chemtrails!

PZ Myers posted about this also.

1 Like

There should certainly be more rigour in reporting what you intended to find, and what you actually find. But the fact that they switched doesn’t imply dishonesty.

A major point of a study is to explore a question you don’t know the answer to. Discovering something that leads to you a different question is an expected outcome.

1 Like

Surely they publish the letter bits in their fraternal journal ‘CONSORT, Please!’ [Puts very small pin in it.]

This topic was automatically closed after 5 days. New replies are no longer allowed.