[tor-talk] mildly updated statistical analysis of Tor metrics data

[Virgil Griffith]

I just realized there's a better way to control for some of these
effects using multivariate models.  Another update will come later.

-V