Although I didn't know it at the time, the first day of the green group happened to be the first day of the EFF Tor Challenge. So that makes some sense.
Secondly, a two-sample t-test put the blue and red groups into different populations. Thirdly, a regression analysis of the solely the red data puts a slope at -1.889e-05 (essentially zero) with a p-value of 10^-11. -V On Thu, Aug 7, 2014 at 10:42 PM, Anders Andersson <pipat...@gmail.com> wrote: > Why do you think they belong to different groups? It's absolutely not > obvious if the colorization is removed. All of your "groups" have a > preceding slope indicating that something started before your limits, > and they seem to be normal variations of a general down slope that > started in 2013. > > On Fri, Aug 8, 2014 at 12:54 AM, Virgil Griffith <i...@virgil.gr> wrote: > > I've been looking through the various historical data from > > metrics.torproject.org. > > > > If you plot the 'used bandwidth' divided by the 'advertised bandwidth' > > (meant to be a rough measure of network congestion), you get three > distinct > > groups, seen here: > > > > http://dl.dropbox.com/u/3308162/three_groups.png > > What happened on the dates 2013-10-09 and 2014-06-06 which could have > > resulted in these striking drops in network congestion? > > > > -V > > -- > > tor-talk mailing list - tor-talk@lists.torproject.org > > To unsubscribe or change other settings go to > > https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-talk > -- > tor-talk mailing list - tor-talk@lists.torproject.org > To unsubscribe or change other settings go to > https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-talk > -- tor-talk mailing list - tor-talk@lists.torproject.org To unsubscribe or change other settings go to https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-talk
