3-D Bar Graph "Masterpiece"

I encountered this post on how to turn a "boring" bar graph into a 3-D "masterpiece." What's striking to me is that most of the people commenting actually want to replicate this graph, even though it violates the basic principles of effective statistical graphics, according to Tufte and others. For example, the 3-D effect distorts the information displayed by the "boring" bar graph, making comparisons difficult, and the visualization effects distract from the underlying data as conveyed by the differing heights of the bars. Here's the "masterpiece" in its full glory:

Violin Plots

Violin plots are an excellent way of displaying the distribution of a continuous variable by levels of a categorical variable. In essence, violin plots are box plots and kernel density plots combined. For instance, here are a set of violin plots from Stata's auto data:

These same data could also be displayed in tabular form, but again this is case in which a graphical display is a more effective way to examine and convey the patterns in the data.

Big Data and the End of Theory?

An article in The Guardian gives appropriate caution to claims that data analysis (and only data analysis) is the solution for all or even most academic and research problems. As Max Weber observed in  his brilliant essay on objectivity in the social sciences, even the process of data analysis depends on values that cannot be empirically proven as right or wrong: "The 'objectivity' of the social sciences depends [..] on the fact that the empirical data are always related to those value-ideas which alone make them worth knowing and the significance of the empirical data is derived from these value-ideas. But these data can never become the foundation for the empirically impossible proof of the validity of the value-ideas."

Era of Big Data

The New York Times has a great article discussing the era of big data. This might have a Kurzweil-esque ring to it, but due to technology change big data is becoming increasingly available and ready for analysis: in fact, there are more data sets out there than brains to analyze them, especially when one notes the incredible number of combinations of analyses that could be conducted even on a single data set with 100 variables (in what is known as the curse of dimensionality). However, one problem with big data is that, since so much of the data are collected by private entities, much of it may not be available to academics and independent researchers.

Theory of Everything?

A new journal in biology called Life has published an unusual article in its inaugural edition: a paper by Erik Andrulis titled the "Theory of the Origin, Evolution, and Nature of Life." You can find the paper here. At 105 pages and 800 references, his paper seems Sokal-like, except it apparently is not a hoax at all. As a result this paper is unusual, but especially so for two reasons: first, Andrulis is apparently a well-respected biological scientist who has done important work on RNA, and second, Life appears to have all the trappings of a well-respected, peer-reviewed scientific journal, including a well-respected editorial board.

In essence, Andrulis outlines a theoretical framework  that (supposedly) unifies the microcosmic and macrocosmic realms, validates predicted laws of nature, and explains the origin and evolution of cellular life. Like most non-biologists encountering this paper, I've only skimmed it, but apparently reality consists of geometric entities "gyres." Sounds good, except Andrulis provides no evidence (as far as I can tell) that these gyres exist.

It's easy to criticize this paper, if only for ambition of his theory. In one section he purports to unify all laws of nature, while in another he addresses the meaning of life. Even more astounding is the offhand way he presents his theory. For example, on page 55, Andrulis briefly remarks: "Please note the unity of reality and life as revealed by this theory." Can the unity of reality even be "noted"? However, my favorite part of the paper is on page 61, simply because of the sheer grandiosity of his assertion: "I refer the reader to the Theory section for a complete presentation of theoretical answers to many of science’s most challenging questions."

Questions abound how this paper was published despite peer review (perhaps it was a publicity stunt for the journal), and about the sanity of Erik Andrulis. From the position of a sociologist of culture, however, the more interesting question concerns why this paper was so heavily criticized, and whether or not papers such as these have a place in scientific journals. Andrulis' paper, I suspect, is filled with flaws and inconsistencies, but I contend there is often insight from theoretical frameworks that we "know" are generally wrong. Thus, the problem, from my perspective, is not that Andrulis wrote this paper, but rather that there is not a biological journal (to my knowledge) where scientists can publish speculative or half-formed theories that are probably "wrong" but nonetheless help us think about the world in a different way. (Sociology, in contrast, in part because of our methodological pluralism and historical connections with philosophy, has a number of journals in which theories, even those that are highly speculative, can be developed and publicized.)

Murray on Cultural Inequality

The conservative sociologist Charles Murray has written a new book on cultural inequality, and he's written about his main arguments here in the Wall Street Journal. There are two glaring problems with his argument, however. First, although I appreciate his attempts to examine cultural factors of the economy, he frequently conflates behaviors with culture (which consist of values, attitudes, beliefs, not behaviors arising from these symbolic constructs). This muddles his argument, and leads to a profusion of of ad hoc claims that are weakly supported by the data, if at all. Second, his
explanation for cultural inequality falls short: in particular, he ignores how lack of public investments and conservative economic policies (for example, lack of investment in public transportation, public spaces, universal welfare systems, and the growth of car-based urban sprawl based on the profit-making concerns of private developers, among other things) are leading causes of the cultural fragmentation he is concerned about.

Inequality versus Dispersion

I'm glad to see that Alan Krueger, chairman of the Council of Economic Advisers (a fancy name for a panel of three economists), discussed the problems with inequality in his address today. You can find his remarks and graphs here. I liked his graphs, and he shows convincingly many of the standard findings in sociology and political science on politics and inequality in the United States. However, I found the following comments puzzling:

Although I have done much research in my career on inequality, I used to have an aversion to using the term inequality. The Wall Street Journal ran an article in the mid-1990s that noted that I prefer to use the term “dispersion.” But the rise in income dispersion – along so many dimensions – has gotten to be so high, that I now think that inequality is a more appropriate term.

The mixing of the statistical concept of dispersion with the sociological concept of inequality muddles the discussion. It's true that any distribution is often described by some measure of dispersion (e.g., standard deviation) and central tendency (e.g., mean or mode). But inequality encompasses a concept of equity, as well as some concept of disparity (or disparities), neither of which is analogous to the statistical concept of dispersion. Moreover, if we use Krueger's logic it's unclear at what threshold "dispersion" is labeled "inequality"; for instance, his comments imply that Sweden currently has dispersion, while the United States has inequality, although many Swedes would probably disagree.

Congratulations to the Digging into Data Recipients

The list of the round two award recipients for the 2011 Digging into Data challenge are listed here.

Upper Class are More Unethical

This is the first study I've seen that explicitly claims that people in higher social classes are more unethical than those from lower classes; furthermore, the authors claim that this tendency is "accounted for, in part, by their more favorable attitudes toward greed." Aside from the well-known problems from the so-called "sophomore problem," this is a compelling and disturbing study that warrants replication in other contexts (based on, for instance, observational data from other populations).

Culturomics and the Google Ngram Viewer

In case you've missed it, I highly recommend exploring the Google Ngram Viewer, which allows for basic analysis of 5.2 million books catalogued by Google. The Ngram Viewer is an exemplar tool for what the authors call "cultromics" or the "the application of high-throughput data collection and analysis to the study of human culture." For example, the Ngram Viewer is useful for examining over-time changes in concepts used by scholars and others. For anyone doubting the decline of the concept of "social class" in the realm of ideas (but of course not in reality), here is a compelling chart (the most reliability data are from 1800 to 2000):

You can also download the entire data set yourself, but beware: the data set is (ahem) large.