2 ms·
Simple rule, less about trust and more about science: the paper must be at least 5 (five) years old and have one (1) independent confirmation before I'll consid
by aristus 13y ago
Simple rule, less about trust and more about science: the paper must be at least 5 (five) years old and have one (1) independent confirmation before I'll consider reading it.
It's possible to read the original papers and look for obvious signs of trouble, eg small sample sizes, improper randomization, etc. Basic science & statistics is not rocket science.
But the independent conformation rule is a really good filter. If, after 5 years, there's no follow up, there's good odds the paper was wrong. Or even if it's right, whatever it claims isn't considered important enough to follow up on. Even in the case of outright fraud like the Lancet autism article, the lack of independent studies reproducing the claimed correlation would have been a red flag.
- fractallyte 13y agoUnfortunately, some very important papers wouldn't make it past this test... The startup I'm working with has developed a unique, revolutionary product. (Unusually, this is a true statement - not marketing.) There are a number of published papers, and 8 year follow-ups. But journals won't even publish the 5 year follow-ups! Why? Because it is a genuine paradigm-changer. We can't get independent confirmation, because no one wants to touch it. It threatens a global multi-billion dollar industry. There's a 'mafia' in almost every profitable industry. Until I joined this startup, I had no notion that this niche was so full of medical practitioners readily sidestepping the Hippocratic Oath for the goal of profit.
- aristus 13y agoThat's fine, as far as John Q Layman's life is concerned, cutting through false positives with minimal effort. There will be false negatives with any rough filter. It sucks for you, but that's your fight to fight.