Weight has too many unknown related variables to ascertain solely from checkin data. Guessing weight is usually dependent on identifying body type and height to a high degree of accuracy. That's how they do it at the county fair, anyway.
Apart from body type and height, metabolism and activity level would be required to track weight changes from checkin to checkin. Did I walk from my house to McDonalds, or drive? How would Foursquare ever know? Hrmmm… Perhaps, using something like the iPhone/Android location data caches, we could get a "transit speed" estimate, plus "route taken" data, then map that to the mode of travel – foot, bike, automobile, train, airplane, etc.
That still leaves out caloric intake/burn and activity level… Maybe I'm jumping rope on the ferry boat from Alameda to SF, or maybe I'm sitting on my ass having one of their fine cocktails and a bag of M&Ms. You could find out, with some combination of credit card purchase history, access to the onboard security cameras, or a device I wear to track motion/activity, like the FitBit.
But… What about metabolism? For that, unfortunately, we really do need a combination of the above data over time, plus weight data for at least some of the same period.
Which kinda breaks the whole thing – you need weight at point A and B, to even start to estimate weight at point C. And yes, if you did have a way to measure body type and height via those security cams, plus an accurate calorie count of foods consumed, that would possibly work. But then, you'd already have so much other data on hand, that the checkins themselves wouldn't add much insight into weight specifically.
I'm vastly oversimplifying things in a way that'll probably make me look like an idiot to an expert in personal data analytics… But, after doing a lot of research in this area (the company name is NaviDate, after all), I've come to the conclusion that checkin data is very valuable as a source of insight – just not about weight. Personality type, income level, or even gender prediction on the other hand… But that's another whole question. 😉