When is something statistically significant really significant?

Is it true that “People born under the astrological sign of Leo are 15% more likely to be admitted to the hospital with gastric bleeding than those born under the other 11 signs.”

Speaking at the American Association for the Advancement of Science, Dr. Peter Austin of the Institute for Clinical Evaluative Sciences in Toronto provides insight into a common problem that reporters (and the rest of us for that matter) should pay more attention to.

I think about it often when selecting studies to summarize here and while working at my paying job.

For example, while working on a paying project earlier this week, a statistician kept reporting that an outcome was significant by one statistic, but not by another.

“Why do we have to use two statistics,” I asked, “isn’t one enough to show a significant difference.” It turns out we needed both in order to balance the shortcomings of the first statistic.

With respect to the goal of this blog, it’s not necessary — or even desirable — to list every significant finding from CAM studies, just the significant ones.

Got that?

Thanks to Instapundit for directing me to this article in the Economist.

3/3/07 08:58 JR