- Hospital costs at one hospital over 200 years (via Marginal Revolution). The graph of hospital costs versus year illustrates what you should never do: make a graph like that using an unlogged (raw) y axis (cost). It is nearly impossible — or actually impossible — to learn anything from the low numbers since they are so close to zero. New England of Journal of Medicine editors need better statistical advice.
- Cooking with Marmite.
- Steve Jobs, orthorexic.
- The Telgi fake stamp paper scam.
- preorder Trust Me I’m Lying by Ryan Holiday
It’s not true that you can’t learn anything from the low numbers. You can get a sense that hospital costs were very low relative to late 20th century costs, and you can get a sense of when hospital costs started climbing rapidly. The effect on the reader — especially one who might not appreciate he is looking at a log plot (which for NEJM readers would be quite a lot I would imagine) wouldn’t be as large with a log plot. So it depends on what the presenter of the graph wants to accomplish. If he wants the reader to see a rise in hospital costs in the 1870s then a fall in the 1910s then a log plot is best. If he just wants to point out the large and dramatic climb of healthcare costs in the late 20th century, I think an unlogged plot is best.
Seth: “You can get a sense that hospital costs were very low relative to late 20th century costs, and you can get a sense of when hospital costs started climbing rapidly.” You may get the wrong sense. With no correction for inflation who knows what the hospital costs in 1910 were relative to late 20th century costs. Nor do you know when hospital costs started climbing rapidly. Perhaps hidden in those tiny values are large percentage increases. Perhaps the sharp rise apparent on the graph is an artifact of changes in the inflation rate.
The cost is in 2010 dollars.
Seth: Good to know. Then I would just say that the graph makes it hard to see anything interesting apart from the date when costs first started to rise quickly.
To me the interesting thing (purely visual estimate) is to see that for an (on average) 2 times decrease of the mortality rate a modern (post-1965) hospital required a 7-fold cost increase.