Friday, September 25, 2009

More Statistics!

Virtually every news network is trumpeting the "breakthrough" in an HIV vaccine trial, citing a "31.2%" reduction in the risk of being infected!

"In the three-year experiment, 74 of 8,198 people who received placebo shots became infected with HIV compared with 51 of 8,197 people who received the vaccine, suggesting the vaccine regimen could have reduced the risk of being infected by 31%.

The NIH said the results are statistically significant."
Statistically significant? Really?

Considering that less than 1% of either group contracted the HIV virus, how on earth can comparing the two be "statistically significant"??? Couldn't virtually any rare occurrence break down with a similar disparity in distribution?

Even if the results are "statistically significant", is it really responsible to conclude the vaccine has resulted in a "31.2% reduction"? That seems a very misleading number to me. Throw in a few isolated cases to one side or the other and the percentage would change significantly.

I am no statistician but it seems to me you might come to such numbers arbitrarily, substituting virtually any rare phenomenon.

Had the study consisted of 80,000 people, then these results might be interesting. But to come to such conclusions with so few actually contracting HIV seems highly, grossly irresponsible.

No comments:

Post a Comment