Water balloons without the balloons
Inside the secret Soviet failed moon program
In-browser compiler for 40 languages
The interruptible programmer
Great depression and the gold standard
How to cheat with graphs
Interviews with 2010 Fields medalists:
Triple-negative breast cancer
The most remote island
Great links, that browser/compiler is pretty slick.
Ok, I don’t understand the article “How to Cheat with Graphs.” The author writes, “Starting at 0 is a well known trick, and a good one.” The truth is quite the opposite. NOT starting at zero is a well known trick. If you don’t start at zero, you have no visual way of determining how steep the increase really is. If you choose your y-axis limits where you want, you can always make any increase, no matter how tiny, appear as large as you want, typically taking up the entire graph. THAT is the “well known” trick. There is no way to compare such graphs visually; you must study the axes. Of course, Briggs was right that it was disingenuous of Krugman to present one graph starting from zero and the other with arbitrary limits. But presenting both graphs with arbitrary limits, as Briggs suggests is standard, would not solve the problem. The graphs are only comparable if you start both from zero.
Ok, Briggs is a statistician and I am not, so perhaps I am missing something. But I can’t for the life of me see what.
Scott: I guess the resolution is that the decision of where to start the axes depends on context. If absolute values are important, start from zero. If changes are important, start from where the action starts. And above all be consistent, as you pointed out, when you’re making a comparison.
John D. Cook
Subscribe by Email