3 ms·

I find this paper rather flawed (article link: https://www.sciencedirect.com/science/article/pii/S0013935118300355?via%3Dihub https://www.sciencedirect.com/scie

I find this paper rather flawed (article link: https://www.sciencedirect.com/science/article/pii/S0013935118300355?via%3Dihub https://www.sciencedirect.com/science/article/pii/S001393511... ). I don't claim to be an expert on EMF effects on the human body, but I interpret his core claim to be that because the seven effect studies he cites had insufficient power to determine there was no correlation, there must be correlation because "similar" EMF studies did show a response, and therefore the effects should be considered established (ergo the fairly clickbait title). I use quotes for similar because it doesn't seem they were all that similar, which is why he objects to the studies he references. If he had simply restricted himself to saying that more research was required because previous studies were inadequate (due to small sample size, methodological issues, etc.), I think this would have been a lot more defensible. Right now it reads like tinfoil hat time.

gbrown 8y ago
I'm also not an expert in this field, but I agree that the conclusions presented here appear at first glance to be much too strong. For one, we should all be highly skeptical of meta analyses to begin with - there is every incentive to report "significant" results, and a documented tendency for such studies to be preferentially published, so the current literature should be enriched with respect to rejections of the null. Also, did I miss it, or is there zero discussion about how the discussed publications were selected? A quick google search turns up lots of additional null results. The authors also do make some correct criticisms of common statistical mistakes. For example, it is true that failure to reject a null hypothesis is not evidence for that null hypothesis in a frequentist framework. Even so, we should consider the power to reject the null, which depends on: 1. The practical or clinical effect size 2. The sample size Even if we do detect statistically significant effects of EMF/WiFi, that in no way implies that the results are of practical significance. The cited studies do have small samples, but no discussion is given to the magnitude of effect which would be expected to have been detected. They also note that previous studies should have published their data and p-values, which is true. My takeaway is that more information is needed, and that we should hear from someone with a good grasp on all the relevant literature to see how representative these studies are. Also, this kind of thing would be clearer if we were all Bayesians, and published our data :P
lambda 8y ago
Also, if you check some of the citations that he lists for dangers of microwave-range radiation (you can click through to the full text of the article), they are things like studies of naval ships, in which the microwave radiation would be at much higher power levels than that which your Wi-Fi router is emitting. This is full-on crank territory here.

gbrown 8y ago

I'm also not an expert in this field, but I agree that the conclusions presented here appear at first glance to be much too strong. For one, we should all be highly skeptical of meta analyses to begin with - there is every incentive to report "significant" results, and a documented tendency for such studies to be preferentially published, so the current literature should be enriched with respect to rejections of the null. Also, did I miss it, or is there zero discussion about how the discussed publications were selected? A quick google search turns up lots of additional null results. The authors also do make some correct criticisms of common statistical mistakes. For example, it is true that failure to reject a null hypothesis is not evidence for that null hypothesis in a frequentist framework. Even so, we should consider the power to reject the null, which depends on: 1. The practical or clinical effect size 2. The sample size Even if we do detect statistically significant effects of EMF/WiFi, that in no way implies that the results are of practical significance. The cited studies do have small samples, but no discussion is given to the magnitude of effect which would be expected to have been detected. They also note that previous studies should have published their data and p-values, which is true. My takeaway is that more information is needed, and that we should hear from someone with a good grasp on all the relevant literature to see how representative these studies are. Also, this kind of thing would be clearer if we were all Bayesians, and published our data :P

lambda 8y ago

Also, if you check some of the citations that he lists for dangers of microwave-range radiation (you can click through to the full text of the article), they are things like studies of naval ships, in which the microwave radiation would be at much higher power levels than that which your Wi-Fi router is emitting. This is full-on crank territory here.