Home | Business News | Browse by Publication | J | Journal of the American Statistical Association

Numerical Issues in Statistical Computing for the Social Scientist.

Article Excerpt
Numerical Issues in Statistical Computing for the Social Scientist.

Micah ALTMAN, Jeff GILL, and Michael P. MCDONALD. Hoboken, NJ: Wiley, 2004. ISBN: 0-471-23633-0. xv + 323 pp. $89.95.

For as long as the social sciences have existed as sciences, they have relied on statisticians to identify appropriate procedures for data analysis and inference. Now that computers have become desktop appliances, the role of the statistician has been relegated primarily to that of a programmer, a producer of the black box into which the social scientist pours her raw data. The black box produces the summary table, and the p values, on which the scientist bases her evaluation of a cognitive model, efficacy of a treatment, theory of consumer behavior, predicted immigration patterns, and so forth. As theories in the social sciences have become more complicated, reflecting the intractability of the complex systems that they are designed to explain, the demands on that black box have grown considerably. Building on a trend started when all-purpose statistical packages and spreadsheets first appeared, procedures like nonlinear regression, maximum likelihood estimation, factor analysis, and Bayesian hierarchical modeling are now performed regularly by many social scientists.

Some packages, like SAS or the open-source project R, have evolved along with the computationally difficult models that social science now prefers to deal with: other packages are less powerful. A great deal can be accomplished with SAS or R, but a significant element of programming will always be required to attain any desired end. To put this review in perspective, it is important to understand my background. I consider myself, as a quantitative psychologist, to be something of a quantitative sophisticate. I have experience programming...

View this article FREE - Now for a Limited Time, try Goliath Business News
Free for 3 Days!