I would argue that certain personality types are more likely to "tweet" and "check in" and therefore cannot be statistically representative of the general popul

I would argue that certain personality types are more likely to "tweet" and "check in" and therefore cannot be statistically representative of the general population. So to base your research on such a dataset could be skewing the result quite significantly.