Re:Graph TV – The Simpsons
by jvdh on Saturday March 29, 2014 @03:12AM
(#7491662)
This is a great example of how to lie with statistics, or what someone can claim by applying just college-level statistics. Those trend lines are completely worthless.
See the trend lines for Futurama: http://graphtv.kevinformatics.com/tt0149460
The difference between all the episodes make having a "trend" very doubtful. Especially season 5 is a wildly varying season where if you take one episode away the line would completely flip.
Much the same can be said for The Next Generation http://graphtv.kevinformatics.com/tt0092455 where you have basically clouds where seemingly at random a line is drawn through it.
Yes, I know there are statistical methods for determining trends, but without data on their accuracy they are pretty much worthless. And you really should use a threshold for those accuracies if you're presenting this kind of data to a very wide audience.
See the trend lines for Futurama: http://graphtv.kevinformatics.com/tt0149460 The difference between all the episodes make having a "trend" very doubtful. Especially season 5 is a wildly varying season where if you take one episode away the line would completely flip.
Much the same can be said for The Next Generation http://graphtv.kevinformatics.com/tt0092455 where you have basically clouds where seemingly at random a line is drawn through it.
Yes, I know there are statistical methods for determining trends, but without data on their accuracy they are pretty much worthless. And you really should use a threshold for those accuracies if you're presenting this kind of data to a very wide audience.