What is the method by which the paper "Chocolate is effective for dieting" fished the media?

ByNicolas Valentin

Institute of Diet and Health(IDH) on March 29, 2015 "Chocolate is effective for dietingIt is calledAnnouncementAs a result, it was posted on various internet media, and it became news until it was broadcast on TV. However, as the author of the paperJohannes BohannonHe revealed that the paper "Chocolate is effective for dieting" revealed that it was magnificent fishing, why he published papers that deceive many people, and surprised that he caught the world media We are disclosing the method.

I Fooled Millions Into Thinking Chocolate Helps Weight Loss. Here's How.

Bohannon, the author of the paper, has a doctorate degree in the beginning, but his specialty field is molecular biology aimed at understanding life phenomena using molecules, and he has not studied about dieting. Also, occupation is not a professor of the university but a journalist who writes scientific articles. According to Mr. Bohannon, a German television producer producing a documentary criticizing the diet industry has contacted him and asked, "I want to verify how easily the experiment on diet becomes the top news in the media" about.

ByAlan Cleaver

Mr. Bohannon is willing to accept requests from television producers, and a magnificent fishing project will be promoted. First of all, we invited participation in a documentary program on diet on Facebook and collected 5 men and 11 women as participants in the experiment. Participants collected were from a wide range of ages from 19 to 67 years old.

Next, when asking an experiment for a practitioner called Mr. Gunter Frank and asking for an experiment, Mr. Frank accepted the request and proposed to further experiment the theme "to make chocolate effective for dieting". When Mr. Bohannon asked Mr. Frank why he decided the experiment theme, "Whole FoodBecause fanatic enthusiasts like chocolate very much. Tasty bitter chocolate should be good for your health. It seems like religion. "

ByElana's pantry

The experiment was started after confirming that nobody had any problems by conducting physical examination for all the participants before the experiment. In the experiment, we divided subjects as "low carbohydrate diet group" and "group to take 42 grams of chocolate everyday while doing low carbohydrate diet" and "group to eat whatever you like" for 3 weeks It is contents to receive. When the experiment is completed, weigh the subjects and collect blood to measure cholesterol level, protein concentration in the blood, etc. Furthermore, we inspected with a total of 18 items such as quality of sleep and health condition.

According to the test results, it turned out that both groups of "low carbohydrate diet group" and "group who ingest 42 grams of chocolate every day while doing low carbohydrate diet" succeeded in dieting about 2.3 kilograms on average. Furthermore, it was revealed that the group who ingested chocolate had been losing about 10% faster than the group not ingesting it, that is, it was able to effectively diet.


The question that emerges here is that "after all it was clear that experiments can diet effectively by eating chocolate?" According to Bohannon, there seems to be a big secret hidden there.

Mr. Bohannon says the secret is "Statistical significance"is about. Statistical significance is a term of statistical meaning "It is hard to believe that it is chance by chance, it is considered to be meaningful". A hypothesis that can be set up to judge whether "hypothesis" is correct or notNull hypothesisAnd the probability that a null hypothesis is established is called "p value". Generally, when the p value is less than "0.05 (5%)" it is said that "data derived from experiments and research is statistically significant".

In the experiment of chocolate, it was found that the probability that the p value becomes 0.05 or less exceeds 60%, and it was found that significant results can be obtained with considerably high probability. However, according to Mr. Bohannon, when examining with a large number of items including a few subjects that are irrelevant to the experiment, the p value becomes less than 0.05 in a plurality of items, so that a significant result is obtained The possibility increases, this is called "p - hacking". Some scientists forcibly set the p-value to 0.05 or less using p-hacking in order to forcibly adapt the established hypothesis and experiment results.

Bohannon says, "For example, the weight of a woman frequently changes more than 3 kg in the menstrual cycle, which is a larger number than our experiment with chocolate diet.It cuts down such accidental elements and gets accurate results In order to get, it is necessary to gather a lot of participants.In addition, participants in the experiment must have a wide range of age ranges, and the number of men and women must be balanced.In our experiment, I did not consider it at all. " In other words, the experiment done by Mr. Bohannon et al., Looking only at the results, it seems that chocolate has promoted dieting, but that the number of subjects is too small to trust.

Mr. Bohannon and his colleagues who are likely to exit easily if the experts verified properly, but when they submitted the results to the journals related to the thesis of 20 articles in a paper, they published from multiple journals in less than 24 hours I got permission. Moreover, it is necessary to verify and evaluate the paperPeer reviewNothing at all. After choosing International Archives of Medicine from journals that received permission to post and paying 600 euros (about 81,000 yen) of the cost of posting, the paper was posted on the International Archives of Medicine two weeks after payment.

After that, Mr. Bohannon made an official website of a fictitious institution called Institute of Diet and Health, and released a press release. After publishing the press release, various web media posted the article "Chocolate is effective for dieting", and it was also introduced on television in the United States and Australia. A few media interviewed Mr. Bohannon, but the content was "Why is chocolate effective for diet?" And "Please give advice for our readers" As one person, there was no one to ask about the numerical value of the experiment result.

Thus Bohannon's grand fishing has been successful, but he says to the reader: "When reading articles about science, the reader will have to learn how to read. If the experiments introduced in the article If you do not disclose the number of subjects, I advise you "Why is not it open?"

The paper that "chocolate is effective for dieting" is said to have been deleted from the International Archives of Medicine.

in Note,   Science, Posted by darkhorse_log