[smgj]

Opprinnelig lagt inn av KarMa40, her.

Min favoritt statistikk-professor har tatt en kikk på tallene bak undersøkelsen om dødelighet og rødt kjøtt: http://healthcorrelator.blogspot.com/

Lenge siden jeg prøvde denslags selv (coeffisienter og bivariat analyse gjør meg svimmel, kjenner jeg), men hans observasjoner om at det ser ut som om inntak av rødt kjøtt ikke øker dødeligheten signifikant for den femtedelen av deltakerne som spiste mest kjøtt, selv ikke for diabetikere er vidunderlig å lese.

Og det går ut ifra at dataene innsamlet er korrekte ... I flg det Chris Kresser sa om metoden i denne studien er ikke det særlig sannsynlig heller. (Lytt gjerne - akkurat den sekvensen er skikkelig morsom.)

Dernest antar man at korellasjon viser casualitet.... Det er heller ikke gitt.

Fra transcript av diskusjonen om datainnhentingmetodene til studien:
http://chriskresser.com/does-red-mea...-risk-of-death

Sitat:

Chris Kresser: OK, so let’s now talk about the data collection methods that were used in this study. But before we get into this, let’s do a little experiment of our own. So, everyone listening to this, take out a piece of paper. If you’re driving, you can skip this step; you can just think about it in your mind. But I want you to write down how many servings of red meat you had last Monday, whenever it is you’re listening to this.
Steve Wright: Ooo, that’s a good one.
Chris Kresser: And then write down exactly how much red meat you had. You know, how many grams of red meat did you have at that serving? OK, now I want you to do the same thing and tell me how much red meat you had on March 15, 2009. How many grams did you eat during that day? And then I want you to tell me how much red meat you had in March of 2008.
Steve Wright: Um, Chris, I don’t know where I was in 2008.
Chris Kresser: Ha-ha, so how’s your list looking right now? Do you feel confident that you can recall how much red meat you ate four years ago and how often you ate it and how many servings and what amount?
Steve Wright: Yeah, I don’t think so.
Chris Kresser: Yeah, I don’t think so either. I don’t think anybody could. And yet that’s the method that was used in this study. It’s called a food frequency questionnaire, and they ask you to write down, to estimate your food consumption, but they only were filled out once every four years.
Steve Wright: Really?
Chris Kresser: Yeah. Every four years.
Steve Wright: And you had to answer questions about the previous four years?
Chris Kresser: You had to answer questions about the previous four years, how many servings of red meat you had, how much you ate, how many servings of vegetables, fruits, dairy products, and other foods, and you know, most people hardly can remember what they ate yesterday, much less a year ago, much less four years ago. And what’s more important is that even if people could remember what they ate accurately, that’s not what they report on these food frequency questionnaires. Instead they report what they should have eaten. And this isn’t just my opinion. This has been scientifically documented over and over again. Researchers have compared the food frequency questionnaires with more accurate diet records where they take food and they meticulously weight it and measure it, and in fact, a validation report like this has been done for both the Health Professionals Follow-up Study and the Nurses’ Health Study, and I’m gonna read you a quote from each of these validation studies. So, from the Health Professionals Follow-up Study, which is one of the data sets that the researchers reported on, the researchers said: “Foods underestimated by the food frequency questionnaires compared with the diet records (i.e., the gold standard) included processed meats, eggs, butter, high-fat dairy products, mayonnaise, creamy salad dressing, refined grains, and sweets and desserts, whereas, most of the vegetable and fruit groups, nuts, high-energy and low-energy drinks, and condiments were overestimated by the food frequency questionnaires.” And then the validation report from the Nurses’ Health Study came to a similar conclusion and said: The “mean daily amounts of each food calculated by the questionnaire and by the dietary record were also compared; the observed differences suggested that responses to the questionnaire tended to overrepresent socially desirable foods.” In other words, this is just basic human psychology. Most people want to look good. We’re flooded with messages about how fat and red meat are bad and how veggies and whole grains are good, so when people come to fill out these questionnaires, they tend to overreport so-called “good for you” foods and underreport so-called “bad for you” foods.