3 ms·
Gone fishing in a fluid trial
- DiabloD3 11y agoThis is a pretty neat paper. It really highlights how much bullshit flows through the scientific paper industry. "Publish or perish" is a bad metric, and it doesn't allow people to ask themselves if their results actually make sense in any realistic sense, but that they will be rewarded if they find some statistical correlation, no matter how inappropriately stupid. Those four guys must have giant balls for publishing this.
- hyperion2010 11y agoOh man, I'm creating slides to teach intro stats to some fellow grad students, this is quite timely. Not quite dead salmon level, but maybe more concerning given the context. Check your assumptions!
- ehudla 11y agoThis is a precursor: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC61047/ http://www.ncbi.nlm.nih.gov/pmc/articles/PMC61047/
- dekhn 11y agoFor those who don't get the joke, the term "fishing" in the paper's title refers to "fishing for significance", a very common practice in academia. Basically, you take a large data set and repeatedly ask different questions of the data. If you ask enough different questions, without correcting for multiple tests, it's possible, purely by chance, to find a P-value (probability of rejecting the null hypothesis incorrectly) that appears significant. One imagines the authors tried each zodiacal sign until they found one that passed the test. One imagines- but cannot say for sure- that a zodiacal sign is not a feature which is truly correlated with the outcome they observed.