3 ms·
I think the dog one ("Mistake: Forcing a relationship by cherry-picking scales") was actually fine for a very specific reason: assuming dogs are of roughly cons
by geofft 7y ago
I think the dog one ("Mistake: Forcing a relationship by cherry-picking scales") was actually fine for a very specific reason: assuming dogs are of roughly constant density, weight is proportional to volume, which has a cubic relationship to a linear dimension like collar size. Therefore, if you have an increase of, say, ~7% (42 to 45) in your linear variable, you want an increase of 1.07^3 in your volumetric variable... which works out as 18 to ~22.1, which would make the two lines overlap even more. You don't want a 7% increase on both scales.
Intuitively, there should be a fairly tight correlation between size and weight, so overlapping them isn't misleading.
- enriquto 7y ago> Intuitively, there should be a fairly tight correlation between size and weight, so overlapping them isn't misleading. Or maybe not, and tiny, thick-necked pugs are becoming more popular and this trend should be reflected somehow!
- geofft 7y agoWouldn't that be reflected in the graph (as rendered originally) if so?
- ukoki 7y agoFor the dog one I would argue that they are so tightly correlated that neck size adds nothing useful to the chart and is just noise. The article is about dogs getting smaller. The weight chart shows this perfectly. The fact that smaller dogs also have smaller necks is neither surprising nor useful.