3 ms·

Whenever there's a study that says X causes Y, there should be a rule that says you have to say by how much. That's a super-important piece of information. The

Whenever there's a study that says X causes Y, there should be a rule that says you have to say by how much. That's a super-important piece of information. The abstract should state it! If a study says that drinking grapefruit juice shortens your lifespan by a year, then I'll cut out grapefruit juice. If it shortens it by a few hours on average, then even if the study is 100% accurate, it's too small to care about. Especially in news articles, we tolerate this imprecision with health-related findings but not in other areas. You don't see the news reporting that "avalanche kills some skiers". They'll at least tell you "avalanche kills 1 skier and second skier still missing" (unfortunate) or "avalanche kills as many as 350 skiers" (major tragedy).

nonbel 9y ago
The problem is that those estimates are conditional on what variables they put into their statistical model, so they are really usually rather meaningless (not that the mere direction is any more reliable, this can switch too). Maybe you can say: "A specific carbonated beverage (batch 1234 from supplier X) provided ad lib via a standard rodent water bottle with sipper tube increased weight gain by 10% over the course of 3 months (as they aged 3-6 months) of male sprague dawley rats bred by Harlan USA in 2015, raised two to a cage, fed Z brand rat chow ad lib (10 % fat content), kept on a 12 hr reversed day/night cycle at 18-23 C with relative humidity maintained between 40-60%. Over the course of the study, the rats were removed from the cage only three times a week: for 10 seconds of weighing in the afternoons MWF."