Comment by apsec112 | Hacker News Reader

Comment by apsec112 | Hacker News Reader

"In the new study, Delp and coauthors compared health data on 42 astronauts that had traveled into space—seven of which got past the magnetosphere and to the Moon—to the medical records of 35 astronauts that were grounded for their careers. The death rate from cardiovascular disease among the Apollo lunar astronauts was a whopping 43 percent, which is around four to five times the rate seen in the non-fliers and low-fliers (nine and 11 percent, respectively)."

Uh... from this paragraph, it seems like the sample size is 7, the expected value is 1/7, and the observed value is 3/7. Is that even statistically significant? Did they adjust for the number of things they tested for? If 3/7 had died of cancer, compared to an expected 1/7, would that have been reported as a health risk too? It seems almost certain that you'd find departures from statistically average causes of death in any group of seven people, since the sample size is so tiny.

It's worse than that ... Expected rate is 2/7 (general population). Observed is 3/7. The who thing is nonsense.

"The researchers found that the mice exposed to radiation, or both radiation and simulated weightlessness, had sustained damage to their blood vessels. Namely, the mice had impaired vasodilation, or problems expanding their blood vessels to adjust for blood pressure. This can be a precursor to heart attacks and stroke. The mice that just experience simulated weightlessness, on the other hand, seemed normal."

This is the main reason why its statistically significant. Sure, mice arent humans. But I say, with confidence, that many rats died to prove these statistics

Exactly. There's simply not enough data to prove or disprove the hypothesis. What's the proposed mechanism? Because if you chose 7 American men born in the 1940s at random and then discovered that 3 had died of heart disease, I don't think anyone would be falling over themselves looking for an explanation. Animal studies on simulated environments would be more useful. As would quantitative work on other astronauts who've experienced intermediate environments for longer periods. (There've been enough astronauts in LEO for statistically significant results).

> Is that even statistically significant?

Not if your significance level is the customary 5%:

    > 1 - pbinom(2, 7, 1/7)
    [1] 0.06522914

The astronauts carried dosimeters, and all got rad doses below what some DOE employees get, so maybe there is better population for an observational study.

The sample appears to contain 42 fliers + 35 non-fliers.

Reply on news.ycombinator.com