3 ms·
This being fivethirtyeight, I wish they could display a bit more statistical sophistication. Okay, so we've got six out of sixteen. That's not a large number,
by hugh4 11y ago
This being fivethirtyeight, I wish they could display a bit more statistical sophistication.
Okay, so we've got six out of sixteen. That's not a large number, but it's not small enough to be meaningless either. Let's talk probabilities. You could have a graph. You could teach the reader some actual statistics. But no.
As far as I can figure out, the writer isn't a statistician at all, so I'm not sure why she's writing for fivethirtyeight.
- ryan_j_naughton 11y agoI think this part of the article directly addresses your complaint: "I asked Andrew Vickers, a statistician at Memorial Sloan Kettering Cancer Center, to take a look at the analysis. He said it wasn’t surprising that there was an association between e-cigarettes and smoking. The real question is how big it is. While this study suggests that the effect is large, Vickers found reason for caution in what statisticians call the “confidence interval,” a plausible range of values for the study results. The confidence intervals here are “ridiculously wide,” he said, which means that the estimate that e-cigarette users are about eight times as likely to take up smoking as non-users is just a rough approximation — the true increase in risk could be anywhere from 30 percent to 5,700 percent."