Via Mendeley, a nice example of several overlapping histograms:
The x axis is overlabelled, but I don’t want to nitpick.
Previous post on histogram visualization: The mythical Gaussian distribution and population differences
Update 12/21/09: JB links to an improved version of the histograms by Eric Drexler below. And Eric links to the data. Thanks!
Damn you, old fogey money hogs!
On the bright side, I'm about to enter the sweet spot…
Very nice graph. The legend should be eliminated, though, and the year labels attached to the curves with arrows.
the version here (adapted from Mendeley's post) has a cleaner x-axis and different style of year-labeling: http://metamodern.com/2009/11/27/great-science-gr…
p.s. hi aleks
I wonder if Andrew would object to using Kernel density estimates here in place of the choppy histograms. There's nothing sancrosanct about using one year bins for the ages (think of what it would look like if the histograms used age in months, rather than years) and the picture would be clearer with a little (not much!) smoothing.
Yes, I agree that directly labeling the lines and smoothing the curves a bit would help. Overall, though, I'm pretty happy with the pattern that I see in the graph. As Andrew Oswald has found, people are at their unhappiest in their forties (search this blog for "Oswald" for the link), so it's only fair that we get the compensating benefit of free NIH money. Or maybe, more plausibly, it's those NIH grants that are giving us so much stress and making us unhappy!
Although the post seems to be focused on design rather than content, I'm worried that the y-axis is NIH grant numbers instead of a per-capita measure.
How do we know that there are few younger PhDs vying for the awards? Can any other blog reader recommend a way to somehow incorporate the per-capita dimension to the chart?
My best idea would be to similarly show the same age distribution of the people who would be potentially competing for these awards – perhaps including distribution of PhDs at all ages or ones in academia.
Great points, Dana. A graph should never be posted without a link to the original data. Then we'd just be able to do it.
Whenever I see something like this I always feel a little off balance because I can't tell if this means something or is just demographics. I mean the Baby Boom people are moving through the ages and the fact that these graphs happen is expected to me. These kinds of graphs (and I've seen others relating to IT jobs) always seem to do things like % of the work force, which is is interesting, but I would be more interested if the graphs were the success percent of each age group in getting grants, or something like that. That would tell me if things are really different.
@ jb —
I’m glad you like the graphical edit that I did for Metamodern.com (though it isn’t up to Tufte standards). I later located NIH statistical data on the ages of new principle investigators vs. time, which can be found in this spreadsheet:
grants.nih.gov/grants/new_investigators/nih_age_data_principal_investigators_1970-2006.xls
————–
If the above is a double-post, please delete.
This definitely looks better! And thanks for the data.
Try correcting for the overall aging of the US population and the data doesn't seem so drastic anymore:
http://dl.dropbox.com/u/118481/tiled.gif
I plotted median age by year using the data linked to:
http://dl.dropbox.com/u/2023750/med-age-NIH-by-ye…
I have never seen a yearly trend for anything that is so linear. Can anyone explain it?
The overall aging of the U.S. population can account for some of the shift of the mean of the distribution, but it can't account for what I find most disturbing: the drop in the number of new principle investigators under age 30 from a substantial level to very nearly zero, and a still-huge ratio of decline up to age 35 or more.
George: The good news is that the slope of the line is less than 1.
Eric: Even if it's explainable by the age profile in the population, a steady increase in age of the grant-holders will have its effects.
It is very interesting and I'd like to know if the methodology for determining NIH grants changed over the time graphed.
I have 'heard' of cases where the better research project was dismissed because the Principal Investigator didn't have sufficient experience compared to older grant applicants.
I'd be interested to see how has the trend impacted on outputs. Percentage of Grant recipients gaining… Patents? Peer review publications? Extension of initial grant? Sorry, I don’t have the answers just more questions. Sign of interesting data I guess.