Economist Gary Becker and Judge Richard Posner have a point-counterpoint blog that focuses on various issues, but most recently raises a few additional issues relating to the Christakis et al Framingham study.
They agree that few of us know our objective obesity (which is defined in body mass index of weight relative to height) and that people who are overweight tend to believe that they are more normal weight than they are. It thus makes sense that ones friends (if they are similar in obesity or thin-ness) help one to shape one’s self-conception of the average weight of society.
Becker points out that the ideal study would look at college freshman randomly assigned roommates and seeing whether those assigned to larger roommates were more likely to put on pounds over the freshman year since that would rule out the motivations of why one chooses friends and unexplained variables.
Becker also points out that the study by Christakis doesn’t account for the social multiplier on weight gain (your friends’ weight gain increases your weight, at some percentage less than 100%. Then in turn your subsequent gain increases their weight. The weight gains from this multiplier decline with each transmission, but still account for weight gains larger than the original weight gain.)