The graph of Roberts’s weight compared to fructose water intake on p. 73 of “What makes food fattening?” is very persuasive in my mind. I don’t think there is any evidence that it is effective in the population at large, but I think it is clear cut that it worked for Roberts.
I don’t think the cynical explanation gets very far. The details of the diet are freely available. There is only a single, cheap, slim book that Roberts published so that someone could learn about the diet in a format other than his website. Roberts could easily be mistaken, but I think his tone has consistently been “here is a little-known, easy technique that was highly effective for me; I have a theory why it could work for you too”. It’s hard to make money by telling someone to take three tablespoons of extra-light olive oil a day in addition to whatever other diet they are following.
One rat is just not statistically significant evidence—especially not when the rat is also the salesman. I don’t know whether Roberts is motivated by wealth, fame, or whatever—nor do I care very much.
Many tests on the same rat can be statistically significant! Do X, Y changes in the rat. Undo it, Y changes back. Repeat until it’s statistically certain connection...
We just have no particular reason to expect that it’ll generalize well to others.
This really stands out to me as a physicist because we do things like one rat tests all the time. Well, usually we get a few other ‘rats’, but we rely heavily on the notion that identically prepared matter is… identical. Biology, of course, doesn’t allow that shortcut.
Clinicians sometimes have a cohort of 1 for rare diseases… but of course that’s simply the best they can do under the circumstances.
Many tests on the same rat can be statistically significant! Do X, Y changes in the rat. Undo it, Y changes back. Repeat until it’s statistically certain connection...
True—but it won’t be too convincing if self-experimenting on yourself with your own diet. Science is based on confirmations of experiments by other scientists.
The graph of Roberts’s weight compared to fructose water intake on p. 73 of “What makes food fattening?” is very persuasive in my mind. I don’t think there is any evidence that it is effective in the population at large, but I think it is clear cut that it worked for Roberts.
I don’t think the cynical explanation gets very far. The details of the diet are freely available. There is only a single, cheap, slim book that Roberts published so that someone could learn about the diet in a format other than his website. Roberts could easily be mistaken, but I think his tone has consistently been “here is a little-known, easy technique that was highly effective for me; I have a theory why it could work for you too”. It’s hard to make money by telling someone to take three tablespoons of extra-light olive oil a day in addition to whatever other diet they are following.
One rat is just not statistically significant evidence—especially not when the rat is also the salesman. I don’t know whether Roberts is motivated by wealth, fame, or whatever—nor do I care very much.
Many tests on the same rat can be statistically significant! Do X, Y changes in the rat. Undo it, Y changes back. Repeat until it’s statistically certain connection...
We just have no particular reason to expect that it’ll generalize well to others.
This really stands out to me as a physicist because we do things like one rat tests all the time. Well, usually we get a few other ‘rats’, but we rely heavily on the notion that identically prepared matter is… identical. Biology, of course, doesn’t allow that shortcut.
Clinicians sometimes have a cohort of 1 for rare diseases… but of course that’s simply the best they can do under the circumstances.
True—but it won’t be too convincing if self-experimenting on yourself with your own diet. Science is based on confirmations of experiments by other scientists.
The rat being the salesman is the more serious issue there, yes.