Issues <5#1>, 5#2, 5#3, and 5#4 form a complete research system for the standard sample: datasets, codebooks, and methods for testing causal hypotheses. This system is being used in Charles McNett's Summer 1989 Computer Workshop, in the NSF 1990-92 Summer Institutes for Comparative Research, and in classrooms and libraries worldwide. The theme of this issue is software solutions for testing causal models with regression analysis, correcting for autocorrelated residuals in spatially or phylogenetically distributed data (Galton's problem). Douglas White, Malcolm Dow and Karl Reitz discuss the methodological results of NSF funded research (Douglas R. White and Michael L. Burton, Principal Investigators, BNS-80-23904) on the effects of autocorrelation in estimating causal models in cross-cultural research. They have written the three programs on this diskette:
AUT-COR (AUToCORrelated Regression),
AUT-SPAT (SPATial AUTocorrelation), and
AUT-TREE (Phylogenetic TREE AUTocorrelation).
Together, these programs are sufficient for testing causal models of cross-cultural hypotheses. The easiest way to use these programs is with files created by the MAPTAB program. The integration between the programs keeps track of variable numbers, coordinates, language phylogeny, and missing data. The latter two programs prepare data for the AUT-COR regression package. In addition, the AUT-SPAT (SPATial AUTocorrelation) program will compute Moran's coefficient of spatial autocorrelation for individual variables.
White's substantive analysis of the spatial organization of cultural systems derives from analysis of Moran's coefficients for all 1240 variables published on the Standard Sample (see issues 1#1, 1#2, 1#4, 2#2, 2#3, 3#1, and 5#4, and summary volumes 5#2 and 5#3), using the AUT-SPAT program.
White also provides a discussion of the various models and assumptions underlying spatial modeling with regression analysis.
Finally, McCleary and Stigler provide documentation of their Poisson regression package, suited for analysis of frequency data in spatial grids, such as archaeological field sites. Their program, with is not immediately appropriate from the current cross-cultural data base, is available to subcribers on request to World Cultures.
Dow, Malcolm M. 1986. Model Selection Procedures for Network Autocorrelated Disturbances Models. Sociological Methods and Research 14(4):403.422.
Dow, Malcolm M., Michael L. Burton, Douglas R. White, and Karl Reitz. 1984. Galton's Problem as Network Autocorrelation. American Ethnologist 11(4):754-770.
McCleary, Richard, and Mark Stiger. 1988. Confirmatory Spatial Modeling by Regressions of a Poisson Variable. Ms.
SUPPLEMENTARY MATERIALS on diskette but NOT INCLUDED ON THIS DISK are also available on request, and may be published in later issues:
Regression Analysis with Autocorrelated Residuals (under revision)
Douglas R. White <b:AUT-COR.ART>
Theory of Distance Decay Processes in Culture (in progress)
Douglas R. White <b:DISDECAY.ART>
Program DICKMARK: Regression of a Poisson Variable
Richard McCleary and Mark Stiger <DICKMARK.DOC>
Execution file: DICKMARK.COM
Updates to earier journal editions are given in the file: <CHANGES.89>