General Disarray

Wired 14.07: What Kind of Genius Are You?: Very interesting article about the studies of David Galenson, an economist at the University of Chicago. The gist of the research is that there are two kinds of innovators: conceptual and experimental. Conceptual innovators make large breakthroughs when they are young, and tend to stagnate later. Experimental innovators work tirelessly to perfect and achieve their breakthroughs, so they take longer to make it big, but tend to produce more influential works later in life.

From the article:

Picasso and Cézanne represent radically different approaches to creation. Picasso thought through his works carefully before he put brush to paper. Like most conceptualists, he figured out in advance what he was trying to create. The underlying idea was what mattered; the rest was mere execution. The hallmark of conceptualists is certainty. They know what they want. And they know when they’ve created it. Cézanne was different. He rarely preconceived a work. He figured out what he was painting by actually painting it. “Picasso signed virtually everything he ever did immediately,” Galenson says. “Cézanne signed less than 10 percent.”

I was trying to figure out which camp I fit in (not that I am or will be an innovator, necessarily, but the basic idea has a lot to do with work habits) when I read this passage. I sound a lot like Cezanne. It isn’t unusual for me to be halfway through a paper before I realize what I really want to say — in which case I end up rewriting the entire thing from a different perspective.

Definitely worth reading, especially if you need a pick-me-up about your career.

A couple of weeks ago there was a controversy when Fox News talking head John Gibson cited statistics showing that a growing proportion of children under 5 in the US are nonwhite, ending his commentary with a directive to go "have more babies." In response to the criticism, he explained that his chief worry was not a changing ethnic makeup, but the tensions that result from immigration, a possible consequence of decreased fertility. I even defended him, since the coverage of his fertility story was in many cases unbalanced — omitting relevant snips of his argument.

But today I read a Newsweek article by Paul Samuleson ("The End of Motherhood," 29 May 2006) that shows that unlike Italy and Germany, the fertility rate in the US is 2.1 children per woman, just over what it takes to replace the population. And the by-ethnicity breakdown shows that it's 1.9 for non-Hispanic Whites, which rules out the possibility that Hispanics are responsible for the sustainable rate.

Best of all, the article ends with this quote from Adam Smith:

[The] chief part of human happiness arises from the consciousness of being beloved

Aw, shucks. And they call it the dismal science…

Lately I’ve been drinking the Kool Aid about data visualization and the graphical presentation of quantitative results. Regression results, for example, are usually presented in a tabular format. There’s nothing wrong with that mode of presentation, but there isn’t anything especially right about it either. My guess would be that it became a standard format because, before the desktop computing revolution, it was much easier to typeset a table than to include a picture. That’s no longer the case. Using Stata, R or even Excel, I can actually make a nice looking graph faster than I can make a pretty table.

Presenting things graphically can make interpretation easier. Of course, it’s important to ensure that you don’t misrepresent the data with your graphic, which is easy to do accidentally in the process of trying to get it to fit on the page. And I think that figures should still be accompanied by detailed tables, which can present data with more accuracy.

But ultimately I think it comes down to an ego thing. I think that people feel that pictures are for the weak. Real men look at tables, not charts, right? Once I got over that same prejudice when it came to mathematics, I noticed a sudden increase in my ability to solve complicated problems. And there’ s no shame in that. The work of Einstein and Feynman is littered with impromptu sketches and graphs.

Here are the pretty pictures I made today (they are regression results showing the midpoint and upper and lower bounds of the 95% confidence interval):

There are a few cool blogs about stats and data visualization. Check out Data Visualization and Visual Culture and Statistical Modeling, Causal Inference and Social Science for a sampling.

This is exactly the kind of stuff that motivates me to blog anonymously. This study finds a high correlation between [high IQ] and [religion less important].
Gene Expression: IQ Religiosity Redux

CiteULike is basically del.icio.us for researchers. Articles are tagged and include all author information, links to related articles, links to other articles by the authors, and links to download pages. Not all of the articles listed can be downloaded for free (many of the links appear to be to Science Direct and similar sites), but most people within academic institutions have subscriptions to the services that host the articles.

 Very cool. I love to see worlds colliding.

First, a cool article on the Social Science Statistics Blog about research tools, that refers to [Stata+Word] as low-tech and [R+LaTeX] as high-tech. As usual, I can’t make up my mind. Somedays I’m all open-source, somedays I like the canned stuff.

Second, a new version of LyX Mac came out recently, and it fixes all of the problems that I was having before (namely inserting math). If you’re not familiar, LyX is a WYSIWYG front-end for LaTeX that can make using LaTeX for serious or even casual typesetting a lot more convenient.

I’m a little torn about “open sourcing,” which is really just transparency, when it comes to research.

I think that good research should be highly transparent. At the most basic level, this means full description of your data and a detailed sketch of your methods and results. These are the baseline standards followed in academia. Beyond those basics, many journals require that the data be submitted, and even if they don’t, many authors will make both data and stats routines available to interested researchers, or even post the materials on line. This is, of course, asking for nitpickers and second guessers, but to some extent, those types should be welcomed.

On the other hand, working for a private consulting firm, I’m frequently under pressure to obscure my methods, specifics about my data, and other aspects of my work that could be considered part of my firm’s intellectual property. This pressure is not because the methods, data, etc. are wrong (at least, not if I can help it), but because disclosure of that information might allow our competitors, or even our clients, to undertake the work that we do on their own. It does not guarantee that they will — part of what we sell is experience and familiarity with the field — but it raises the possibility.

When I’m writing a paper, I don’t mind withholding some specific background information, because that type of information is usually not necessary to understand the work. Withholding this type of information is somewhat common, even in academic work, in order to preserve confidentiality. But when I’m asked to withold a methodological writeup or fit statistics, I get a little uncomfortable. If I can’t report specific parameter estimates, sample sizes, etc., readers have a more difficult time understanding the implications of the analysis that I’ve done. Even though I’m reporting accurate information, I feel like a used car salesman, as though I’m saying “just trust me, the coefficients were significant.”

I completely understand why we might want to keep our practices a secret. They are how we make money. Sometimes I wonder whether we’re trying to protect trade secrets or trying to hide the fact that some of our analytical techniques are very basic and could be used by anyone. But either way, it is understandable.

The same thing troubles me about corporate research, on medicine or technology for example. Yes, those companies make money by doing research, but isn’t it possible that the research would be furthered by a greater degree of transparency or information sharing? Or would that kill all profit margins, causing research to stand still? It’s an open question.

I just wish that my firm would take a more balanced approach. For clients, keep everything confidential and under wraps. But ocassionally, let me write a paper where I get to show off something innovative and interesting, without having to preface everything with “just trust me.”

Economist Allin Cottrell is an amazing guy. He wrote GRETL, an open-source, graphical, cross-platform econometrics package. He’s also the author of “Word Processors: Stupid and Inefficient,” a widely disseminated article in the LaTeX community. In the article, he essentially makes the case for using LaTeX over word. A case with which I almost whole-heartedly agree. The crux of his argument is that in LaTeX you compose — write the words and structure of your document — while in Word you typeset — fidget with formatting, etc.

It is right to prefer composition to typesetting if you are a serious author. Yes, typesetting matters because a readable document is more likely to be read completely. But form should really be an afterthought.

He acknowledges in his article that one can use the composition mode of writing in Word; the problem is that few do. Since Word is probably the most common format for collaborative writing, however, it pays to be able to use Word like a pro. And if you prefer composing to typesetting, then you need to learn to use Word like it was LaTeX. Here’s how:

Use the styles that are built in to word. When you want to insert your title, use the title style. When you want to insert a top-level header, use the Header 1 style, etc. Customize these styles to your liking, then save a blank document containing these styles as a Word template. That way you only have to mess with the typesetting component once. Making custom Word templates is actually much easier than creating custom style files for LaTeX (at least in my opinion).

LaTeX does a great job of formatting papers to look like journal articles. Many journals use LaTeX to make their articles, so this isn’t surprising. The default Word files aren’t so professional. These tips will help you make your document look more like the journals:

• Change the fonts for the header styles to Times New Roman. To do this, open the style palette, select the style you want to change, and click on modify. From this dialog, you can change the font face, size, weight, etc.
• In the same dialog, you can turn on automatic numbering. From the Format dropdown menu, select numbering, then select outline numbering. There are several styles to choose from (1, 1.1, 1.11; 1, 1.A, 1.A.1, etc.)
• Tell the header styles to change to normal paragraph styles upon a carriage return by setting “Style for following paragraph” to normal.
• Tell the normal style to change to a normal indented paragraph upon a carriage return by setting “Style for following paragraph” to normal indented. This way, the first paragraph after a header will not be indented, but every subsequent paragraph will, similar to the style in most books and journals.
• Set your table styles by selecting the Table Grid style. Within the style dialog, there is a dropdown menu named “Apply formatting to.” You can use this menu to change the formatting for header rows, left-most columns, etc. To apply this style, you must select your entire table, not just the next in the table, then click on the Table Grid style.
• Optionally, decrease the line width by changing the margins. Notice that most journals use much larger margins than Word. This is partially because reading long lines is tiring to the human eye.

These are just a few general tips that will help you have (a) more control over the formatting of your document and (b) a more automated writing workflow, so you’ll be able to focus on writing and not formatting, after the initial template is created. Learning how to use Word’s styles well is a good idea; you can format elements well beyond the ones outlined above. If you know how to customize styles, you’ll be able to quickly make global changes to your template, and never have to worry about manually formatting a long document again.

A reason to reconsider money as the source of creativity (and as a consequence, the importance of highly restrictive intellectual property rights regimes):

Studies Find Reward Often No Motivator - GNU Project - Free Software Foundation (FSF)

There is an old joke that nicely illustrates the principle. An elderly man, harassed by the taunts of neighborhood children, finally devises a scheme. He offered to pay each child a dollar if they would all return Tuesday and yell their insults again. They did so eagerly and received the money, but he told them he could only pay 25 cents on Wednesday. When they returned, insulted him again and collected their quarters, he informed them that Thursday’s rate would be just a penny. “Forget it,” they said - and never taunted him again.

This is from on interesting article on GNU.org about rewards and motivation. According to the article, intrinsic motivators can be far more important than extrinsic motivators (like money, prizes, etc.)