Summer Fiction
by our literary editor St. Clair Carr
One of the chief characteristics of the arts is that they are always behind the times. The artistic still take Freud seriously, for example.
Literary fiction is as outmoded as the other arts. We can see that by looking at a couple of examples of recent scientific fiction (not science fiction, but fictional scientific work) and seeing how much more literarily accomplished they are than self-styled literary work.
The terms "scientific fiction" and "fictional scientific work" may surprise you. Thanks to the revolution in scientific thinking brought about by Sir Karl Popper, science has broken the arts' monopoly on fiction. Aware that scientific progress depends on conjecture and disputation, scientific thinkers have started producing fictional tours de force which would leave contemporary writers of literary fiction green with envy, if they could understand them.
In this article we will look at two recent examples of the scientific fiction genre. Although these books are about a field officially considered boring, their intellectual and literary audacity makes them far more interesting than most current novels (and perhaps all).
These two books are Statistical Power Analysis: A Simple and General Model for Traditional and Modern Hypothesis Testing by Kevin R. Murphy and Brett Myors (Erlbaum, 1998), and Missing Data by Paul D. Allison (Sage, 2002).
Murphy and Myors' book is at bottom a fantasy. However, this fantasy ends up improving our understanding of the real world – not our belief that we understand the real world, which is all that literary fiction usually ends up improving, but our actual understanding of it.
For those of you unacquainted with the fascinating field of statistical power analysis, it concerns itself with the relationship between the size of the sample one studies and the size of the effects one can detect with statistical tests. If you are comparing two groups' scores on a test for example, you want a sample that is big enough to give you a reasonable chance of detecting a difference between your groups but not so large that your tests conclude that even trivial differences between the groups are significant. Implicit in this analysis is the assumption that the scores of the two groups may not differ.
Murphy and Myors' brilliant little treatise is based on the idea that this assumption is false. Groups always differ, they claim. That's right – no two classes ever get the same test scores, no two hospitals ever have the same recovery rates, no therapy ever fails to work, and so on.
You don't have to think about that idea very long to realize that it's wrong. However, the statistical community is willing to suspend disbelief, and Murphy and Myors take advantage of that willingness to spin out a detailed analysis founded on this flawed foundation. As followers of Sir Karl will not be surprised to learn, the result is actually useful.
Murphy and Myors end up reformulating the whole enterprise of scientific research so that it is more efficient and more effective. Now, there is probably no one who will agree with their reformulation – this is science, after all, and scientists live to disagree with other scientists – but Murphy and Myors refocus attention on the crucial issues in power analysis. Their book is both an intellectual and a literary accomplishment, and there are few novels you can say that about.
Allison's book deals with the common problem in analysis of how to cope with the typical failure of information to be complete. Anybody who's worked with data is familiar with this problem – people skip some of the questions on questionnaires (especially those about their ages, incomes, and races), Statistics Canada suppresses information for certain census tracts to protect privacy, data are lost through technical failure, and so on.
Allison reviews conventional approaches to this problem, and then describes in detail the trendy new approach, multiple imputation of missing values. As you wade through his dense description of this approach, though, you experience a stunning epiphany – what Allison is recommending is that you fabricate data!
One of the best safeguards against researchers fabricating data has traditionally been that it has been almost impossible to fabricate them without leaving obvious evidence that you have – evdience which is obvious to other researchers, at any rate. What Allison provides is a set of procedures which help you get rid of that evidence. In particular – and this is a crucial point – no longer will your phony data lead to phony conclusions.
Pretty cool, eh? Phony data which lead to valid conclusions. When was the last time a novelist came up with an idea that good? Again, Allison's book is going to provoke disagreement, but again it refocusses attention on crucial issues.
And when was the last time a novelist wrote as effectively as the authors of either of these two books? Each proceeds at a clip the average highly respected novelist would find breathtaking. That's because the authors know what their books are about. They know what they want to say. These last two characteristics are conspicuously lacking among the novel-writing community.
So in satisfying your summer fiction needs, forget about all those "important" literary novels which you're going to forget as soon as you've read them because when all is said and done they tell you exactly nothing. Get your hands on a few statistics books and prepare to have your minds blown.
Summer Fiction © John FitzGerald, 2002
Click the banner or click here for ACTUAL ANALYSIS
Reviews | Home