The “added stats” on the ad Tom mentions below say that 44 women trialled, about 30% lost an english dress size and a total of 65% had some positive effect*. Wow, only 35% of the group experienced size gain! So overall 15% did better than doing nothing at all.
Only we don’t know that other factors weren’t controlled. Could it be that people trying a “body sculpting” product perhaps be more aware of their diet over the trial period? And in the end, a population of 44, is this statistically significant? I’m sure other PBS’ers have a stronger grasp of stats, but I seem to recall that at a population of 44 the standard error of the mean is 44–0.5 x the standard deviation, which is approx 0.15 x s. And as s in a bell distribution covers 68% of the data, it looks to me like the standard error of the mean is of a similar magnitude to the number of people who did better than doing nothing at all.
*I might be a little off in those numbers – I didn’t jot them down in my advert-pedant notebook – but they’re about right.