4 ms·
This seems like a noisy low-probability event. > While the number of recorded carcasses increased at the control turbines (7 vs. 18), they decreased at the tre
by 3JPLW 6y ago
This seems like a noisy low-probability event.
> While the number of recorded carcasses increased at the control turbines (7 vs. 18), they decreased at the treated turbines (11 vs. 6 [expected: 28]) (Table 1).
I didn't read the article very closely, but it also reads like they're doing lots of hypothesis tests on different groupings to find significant (p < 0.05) results:
> However, the annual fatality rates fluctuated considerably between years (Figure 3 lower panel), stressing the necessity of a long‐term study. Seasonally, fatality rates (across years) were strongly reduced at the painted turbines after treatment during spring and autumn, but increased during summer (Figure 4). When grouping data by season instead of years, painting reduced seasonal fatality rates by 70.9% (95% CI: 61.7%–77.7%; z = −2.003, p = .042, n = 64).
I'm not convinced.
- TheSoftwareGuy 6y agoI agree with you. For some reason, every now and again studies like this -- ones that are interesting but under-researched -- are posted on HN, and I don't get why. People then treat these studies as if they are conclusive evidence of something. Luckily nothing ever comes of it, but still
- dylan604 6y agomaybe the results of their test will spur some doubting Thomases to pick up the mantle and have a go themselves. They'll either find supporting results or disagreeing results. Either way, sounds like good science. If they are taking the results of the single study to say "This is the final answer", then that would be bad. If they are just saying "These are the results we achieved", then that sounds like what a study is suppose to do.
- 3JPLW 6y agoTo be quite frank, repeated hypothesis tests on ad-hoc groupings is _terrible_ science.
- specialist 6y agoI regard this as another pilot program, to be scaled up. Gotta publish to get more funding.