Via infosthetics, I came across a new and very nice web application for data analysis, Verifiable. Among their featured graphs, there’s a very nice one displaying the association between politics and religion:
This graph also shows how the often-hated bar charts can be effective. In all, the graphs coming out of Verifiable look like some of the best I’ve come across. Previously, I’ve written about ManyEyes, which is quite versatile and allows many data types, and Swivel, which was among the first. Nicely done.
[Several commenters have pointed out (thanks!) that the selection of colors is not good, and that some religions in the list are very similar, or too small to be interesting. When it comes to selecting good colors, I stand by ColorBrewer.]
This bar chart isn't entirely bad. And I happen to have no predisposition against bar charts.
That said, I do have a bit of a problem with this one. One could interpret it as representing a "left-right" (or "right-left") continuum. But including the "other" category greatly skews such a visual representation. Someone skimming the graph may notice, for instance, that the Jehovah's Witnesses have a very small amount of red and blue on the left end and assume that this is an adequate representation of their political leanings in that direction. But all of that brown in the right tail suggests we really can't know what they think.
Pity about the colours though – and they can't be chosen by the user.
Whoever made that graph is extremely confused. "Mormon" is what you call a member of the Church of Jesus Christ of Latter Day Saints. They are the same.
JMP has a "share chart" feature that produces almost exactly the same graphic displayed in this blog post. JMP also has an alternative "frequency chart" option that is similar to the share chart, but configures the display of this same data as 5 side-by-side column bar graphs, which is a more informative display of this particular data.
I agree that (a) the graph is nice, (b) the colors suck, (c) they messed up with Mormons, (d) the Other category should be combined with or plotted next to the Independent category–or maybe plotted separately on the side, and (e) the religious groups should be ordered by R/(R+D) rather than by R alone. A common mistake, by the way.
Overall, though, my judgment is highly positive.
Well, it's the LDS (Church of Jesus Christ of Latter-Day Saints) position that "Mormon" and "LDS" are the same thing, but other groups that share origins with the LDS church disagree. See the Wikipedia page on "Mormon" for more detail. So there are people who might self-identify as Mormon who are not LDS, including members of a number of polygamist/fundamentalist groups, which probably goes a long way to explaining why the LDS church insists it is the only group properly described as "Mormon".
Having said all that, all of the non-LDS groups combined are quite a bit smaller than the LDS church, and it's not clear to me why they would be broken out as a separate category rather than combined either with the LDS or the miscellaneous "other faiths" category. It's rather puzzling.
I do like the graph, too.
Mike's point about "Other" category and Andrew's point about plotting it on the other side not to mess up the proportions in other categories are very good.
It seems that it's more of a missing data situation than an informative category. The actual amount of "Other" responses isn't that interesting, so we could have rendered it as a single grayscale box – the darker the higher the non-response.
A more general comment for using color graphs: something like 5-10% of males are, like me, color blind. So even if they are often "nice", for a non-neglible part of the population they are often a very poor way of presenting results.
Especially important to avoid using red and green (most common type of color blindness).
This graphic bothers me. Apart from the weaknesses already pointed out by others, the dataset is not exactly ideal. All people appear at least twice (in Total Population and their own group), but some appear three times (Total Protestants and Unaffiliated are aggregations of other groups). Aggregations should always be clearly distinguished, they are not here. The Total Protestants and Unaffiliateds are simple sums of their subgroups as they should be, but I have not been able to come up with an addition to match Total Population. I'd be grateful if someone could show me how.
The plot could be improved by drawing the bars to reflect the sizes of the populations involved. These range from 118 to 9472, but the plot gives them all the same weight.
So many published plots make me think of cheap marketing: they are attractive at first glance, but then we see there is something not quite right about them. Our standards should be higher (and Aleks has picked a nice example to get our attention).
Are "Orthodox" and "Reform", Jewish or Christian or some combination?!
As the author of the chart in question, it's neat to read this discussion and criticism. (Someone cares!)
I chose a stacked bar chart after experimenting with a variety of visualizations and judging that this resulted in the densest, clearest picture. There may be better visualizations, and I'd love to see someone make one. (And if our tool can't do it, that'll certainly help us prioritize our development plans.)
Regarding the choice of colors, you're of course completely right. It's a limitation right now that you can't select them — I felt lucky just to get Republican to map to red. I'd prefer to use the traditional party colors, with lighter shades for leaning. We hope to address this soon.
As for where to put "other", I thought carefully about this. I think it would be misleading to place "other" in the middle of the right/left continuum, and likewise would be misleading to omit, so I chose to include it at the end. To the extent that some unknown subset of "other" belongs between Democrats and Republicans, I'm failing to portray that, but it's a tradeoff I thought made sense.
You can copy & edit my chart to put "other" in the middle (just add & remove the Y variables until they're in the right order). I don't think it's quite as clear, but YMMV. You could also argue that there should be _no_ sorting, on principle, since the right/left continuum is somewhat bogus to begin with — but while agreeing with the sentiment, I think in practice it makes the picture harder to see.
The remaining issues mostly stem from the source data (which comes from the New York Times, who in turn credit the Pew folks, but don't link anywhere — which is a bummer). There's obviously text that should have accompanied the data, given the unexplained asterisk. But you've got to make do with what you've got.
Other people have commented on the LDS/Mormon issue in the NYT chart comments, but there's been no response from the Times. Also, I'm no expert, but aren't there people outside LDS who identify as Mormon, even though they may be few in number? I could add more annotation to the chart to clarify the issue, but instead I've made a new version with LDS removed, since the issue is not central to the chart.
If anyone can point me to a better data set, however, or a public link to the primary survey data the NYT republished, I'd love to use it. In fact, comparing such data would be interesting in its own right.
Please keep the feedback coming — I'd really like to make the best possible visualizations, and a better visualization tool, so this is extraordinarily valuable.
In addition to the Mormon screw-up, you also screwed up with the Orthodoxes. What is "Orthodox" – Eastern Christian Orthodoxy, or Jewish Orthodoxy? And does "Greek Orthodox" include the much larger Eastern Orthodox churches (like the Russian, Bulgarian, and Romanian churches)? Very confusing.
For anyone still following this, we recently added support for assigning custom colors to each data series, so I was able to make an updated version of the chart using traditional party colors:
http://verifiable.com/charts/3580
As for Stephen Smith's comment immediately above, I can only reiterate my earlier explanation that the data is not mine, and that I agree that the religious categories are not as clear as they should be. If you know of a superior data set, I would be delighted to use it instead!
Peter, your new colors look great!