If it’s a computed trend-line rather than something someone eyeballed then in my book that is a fitted curve. Anyway, that makes sense; presumably it goes below zero somewhere a little to the left of where it stops. Given the obvious discontinuity, it might have made more sense to plot separate lines for before and after...
It’s computed, yes. And I defer to your grasp of the terminology!
And yes, separate trend-lines for before and after the red vertical line might have been good; if the graph’s creator makes such a version, I will post it.
Hmm, the shift there seems mostly the cause of the changes to self-upvoting, as well as the increase in average karma weight. I think you might be able to adjust for that a bit by just using the vote count field (though obviously that ignores the difference between upvotes and downvotes). You could also resolve every individual anonymized vote and then count the upvotes to properly adjust for this, but that might be a bit of a pain.
Thanks to Betawolf on IRC for this graph, which plots ratio of karma to comment count vs. time, and shows some interesting statistics:
That fitted curve looks pretty dubious in its earlier parts. (Maybe I’m misunderstanding and it isn’t a fitted curve at all?)
It isn’t a fitted curve, no. It’s just a trend-line, which looks curved due to the log y-axis.
If it’s a computed trend-line rather than something someone eyeballed then in my book that is a fitted curve. Anyway, that makes sense; presumably it goes below zero somewhere a little to the left of where it stops. Given the obvious discontinuity, it might have made more sense to plot separate lines for before and after...
It’s computed, yes. And I defer to your grasp of the terminology!
And yes, separate trend-lines for before and after the red vertical line might have been good; if the graph’s creator makes such a version, I will post it.
Update: Here is is another graph (also from Betawolf on IRC) that may address your questions/concerns:
And the same thing on a log scale:
Hmm, the shift there seems mostly the cause of the changes to self-upvoting, as well as the increase in average karma weight. I think you might be able to adjust for that a bit by just using the vote count field (though obviously that ignores the difference between upvotes and downvotes). You could also resolve every individual anonymized vote and then count the upvotes to properly adjust for this, but that might be a bit of a pain.