[conservatism_IS_compassion]

The example was a graph showing the reliability of Chevy trucks seemed to show that Chevy is twice as reliable as Toyota and several times more than Nissan—but the graph really shows that they are all better than 95% reliable, and there is only a ~3% difference in reliability between the most and least reliable. But then, at the end, the narrator explains that an ocean temperature graph showing little change since 1880 is misleading because it *didn’t* change the scale to exaggerate the change.

If you want to determine how graphing techniques can be manipulated, you have to apply the same standards across the board.

Yes and no. On face value, you are absolutely correct. But then, arguably, you would always express/plot temperature in degrees Kelvin or degrees Rankine. That is, referenced to absolute zero based on perfect gas theory. Which has bupkis to do with whether you can grow bananas.

So rather than saying that the scale of a graph imparts a “bias," it is the fact that any scale explicitly or implicitly expresses a perspective.

Just try to plot the price of a stock without biasing your estimate of the future price.

[exDemMom]

I have never heard of degrees Rankine before. And for most purposes, it makes no difference whether I plot temperature differences in degrees Celsius or Kelvin—the only change would be the magnitude of the number on the y axis.

I love to make graphs. When done correctly, they convey a lot of information. You have to be honest about the graphs, though—if you are graphing two similar data sets and using different scales, you need to point that out very clearly. Just as you need to state all of the assumptions used in creating the graph.

I’ve never graphed stock prices, only scientific stuff.