I appreciate the effort, but I fear that the message that many have taken from Tufte is “graphs should be cool” rather than “graphs should be clear.” As Yu-Sung put it, “I am still figuring out how to read it.”
I appreciate the effort, but I fear that the message that many have taken from Tufte is “graphs should be cool” rather than “graphs should be clear.” As Yu-Sung put it, “I am still figuring out how to read it.”
Actually, I found the graphics in that article pretty easy to figure out. I don't think they were going for "cool", they took a number of subsets of the data, and presented the subsets in a straightforward way, using "small multiples", but also without a lot of clutter.
Jon,
The #1 thing I get from this graph is that there are more white people than black or hispanic people. Beyond that, I agree that they're conveying a lot of information, but I'm just too distracted by the Rohrschach patterns.
I like the concept, but I feel that a few points could have been handled better:
The area vs. curve aspect is quite distracting. If both were partially transparent shaded regions or distinguishable curves, it would be much more readable.
I think that it would be better to display the data as proportions, rather than totals.
I wonder if it's really necessary to show the entire age profile, or if 5 or 10 year bins would have sufficed. If the latter is true, a simple bar plot would have worked
I wonder if separating the male & female data a bit more would have been better. Separate plots or stacking (non-inverted) plots would have been more readable, I feel.
I actually think those graphs are excellent. The more you look at them, the more that becomes apparent. And that's because of the richness, not the unnecessary complexity.
I had a similar experience in the graph of Napolean's Russian campaign that Tufte made famous.
I'm pleased that they took the effort to try to show everything. Like Andrew, I wish they had normalised the admissions as a proportion of racial category, instead of just showing that "there are more white people".
I'm also puzzled by the absence of a scale below 40. I would have liked it to go down to 30 and 20, especially to see if some of those younger peaks in 1996 were the same birth group 20 years on in the 2005 peaks. Or perhaps the curves should have been on date of birth and not age at admission?
Interesting that marijuana was a drug for which lots of teenagers were checked in (forced to check in?), and twenty years on, it's still teenagers. For other drugs, the bump has moved on more or less year for year.
Aside, I rather like that drug-abuse expert and column writer Charles M. Blow is another dentist named Dennis.
– the drug waves seem to move on. It's the same people doing smoked cocaine, only 9 years later.
– substance abuse is a problem for minority females OR addicted minority males do not enter rehab. Why? Machoism? Are they sent straight to prison and are not given a chance?
– the minority graphs should have been scale by the population proportion.
The only shortcoming I found in the charts was the lack of scale gridlines below age 40, as Derek pointed out.
I agree somewhat with AC Thomas about the waves, but it is not correct to conclude "It's the same people". All you can say is that the same generation did things, shifted in time.
The marijuanna is an exception: That looks like young people's sport, although there is more area a little bit farther up in 2005, looking like some people kept the habit.
Minority graphs could have been scaled by population, but (a) that would exaggerate variability, (b) might have been mistaken as an argument for exaggerating behavior by race, and (c) would confound the presentation with a time-varying proportion of a race, thus withdrawing the ability to do things like examine relative frequency of intakes.
It seems to me that the major failing in this case is that the textual interpretation does not match the figure.
The possibility of cohort-dependent waves in drug use mentioned above is at odds with the treating the increase in admissions for older women as a "disturbing trend", since it implies that drug use for some drugs is also decreasing for younger people and that the trend for older women is similar to that for older men.
The column is about differences between males and females, but the figure highlights racial, drug, and age differences.
I like these plots a lot and find them pretty easy to read and get stuff out of; I spent several minutes looking through them in the paper. I agree with the people who are saying it would be better to scale the racial categories by population (or perhaps by population of the relevant age group, if that makes a difference). The 40-year-old age line is too prominent, and there should be some kind of indicator for 20 and 30 years old, too, but that's a quibble.