Hacker News Reader

Comment by avian | Hacker News Reader

Just note that this friendly graph employs the bad practice of not having the vertical axis origin at 0, which visually exaggerates the change.

View on HN

I disagree. It's a multiplicative factor vertically centered to 1. Seems better to me than if it had been aligned to 0.

Not necessarily a bad practice. Most of the graphs I've seen use this technique, do so to make the changes in value visible, not to deceive the viewer.

That's just it: if the changes are only visible if you chop off the bottom of the graph, the probability they aren't significant increases.

How would you feel if your body temperature increased 3%? Barely visible in a 0-based graph.

What is a 3% change in body temperature? From absolute zero?

(My point is that temperature is probably an odd man out in graph terms, our usual reference points are not actually at zero.)

If the change is invisible on a 0-based graph, maybe the change isn’t so important....

This one only exaggerates things by a factor of 2, at least.

The cosmic microwave background is invisible on a 0-based image.

Only if you’re trying to show the CMB and galaxies in the same image. Since doing so would be misleading as to the brightness of the CMB, that seems like a feature, not a bug.

No, if you insist on a uniform colormap that includes 0K (or whatever measurement you're showing the CMB with) then there's absolutely no way to make the CMB visible. The average temperature is 2.72548 K with variations on the order of 0.00057 K. Its incredible uniformity is one of its more important features.

Oh, you meant variations in the CMB. I thought you just meant making the CMB visible.

0 is only relevant for linear utility/impact functions The effect of an isotope isn't necessarily linear in its magnitude from 0. The baseline should be something that matters, like maximum safe dose.

If you have a baseline and a scale that indicates an actual effect, then by all means use it. That’s not the case here, nor has it been the case for any other such graph I can remember seeing. Universally, the bottom is chosen as the minimum value in the data being plotted.

There are many important changes that are hard to see on a 0-based graph.

If you plot human body temperature in Kelvins, the difference between someone with hypothermia and someone with a raging fever are relatively slight (under 2%).

Doubling is pretty significant, as is the rate of increase prior to, and dropoff after, 1963.

If some sort of catastrophe befalls our current civilization and future archaeologists attempt to C14 date our fossils and artifacts derived from carbon sequestered during this era, their results will be wildly unreliable.

Climate change is significant, even though I'm certain that temperature charted against 0 Kelvin isn't too visible.

Zero can be just as arbitrary a starting point as non-zero.

Reply on news.ycombinator.com