# Causal graphs from cross-cultural research

Cross-cultural causality project -- SCCS R package

## Contents

- 1 New references
- 2 Draft article
- 3 Invitation
- 4 Article sketches
- 5 Sketch for Notre Dame Undergrads talk March 2010
- 6 Sketch of the Causal graphs comparative ethnography project

## New references

- Holden, C. J. and Mace, R. 2005. The cow is the enemy of matriliny' using phylogenetic methods to investigate cultural evolution in Africa. In:
- Mace, R., Holden, C. J. and Shennan, S. 2005. The Evolution of Cultural Diversity: a Phylogenetic Approach. London: UCL Press.

•Holden, C. and Mace, R. (2003). Spread of cattle led to the loss of matriliny in Africa: a co-evolutionary analysis. Proceedings of the Royal Society B. Vol 270. 2425-2433.

## Draft article

- Cultural Consequences of Regionally Fluctuating Inequality see: Exploratory causal analysis for networks of ethnographically well-studied populations

## Invitation

Invitation from J. of Intelligent Learning Systems and Applications From: jilsa@gremail.org <jilsa@gremail.org>

Dear Dr. Douglas R. White,

I am writing to invite you to submit an article to the Journal of Intelligent Learning Systems and Applications. With an open access publication model of this journal, all interested readers around the world can freely access articles online at http://www.scirp.org/journal/jilsa without subscription.

Editor, JILSA

## Article sketches

### Sketch of a Journal of Intelligent Learning Systems and Applications --JILSA article

"Meta-analysis of Causal Graphs at the Societal Level"

### Sketch of an AJS article

"Endogenous and Causal Networks among Cross-Cultural Variables"

- 1 Charles F. Manski. 1995. Identification Problems in the Social Sciences. Harvard: Harvard University Press.

- Identification and Endogenous Variables

- 2

Possible name change for the new fall class "Causal Networks among World Cultural Variables"

### Sociological Methodology and Research draft

Indep/Depvar list - All see graphs

- Human Social Complexity - Causal Networks among World Cultures fall 2010
- Exploratory Causal Analysis within and among sociocultural groups

- Polygyny

### Sketch of an American Anthropologist paper

The Anthropology of Causality within and between indigenous cultures

#### Causality, Morality, Inequality, Envy, and Anger in Human Exchange Systems

#### Cross-Cultural Consequences of Regionally Fluctuating Inequality

### Sketch of a Ethnology paper

- Post-Ethnography of Pul Eliya

- The focus here would be on the consanguineally intermarried segment (CIS) of the village and an adjoining segment in another village, as distinct from the "outside" marriages. Then to trace the CIS in terms of land and other disputes, and how these play out differently among "outsiders."

### Sketch of a Human Organization paper

- Gwembe Tonga - causality discovery for the Thayer Scudder project

## Sketch for Notre Dame Undergrads talk March 2010

(Actual talk was on ethnographic network analysis)

G. P. Murdock and I developed the standard cross-cultural sample (SCCS) in l969 with the purpose of establishing a cumulative database for scientific investigation of the variability in human cultures. It is the gold standard in the field, with ethnographic sources inventoried and available on 186 ethnographic cases pinpointed in space in time so that the information is spatiotemporally resolved to permit further investigation. Hundreds of research projects by different authors have coded thousands of variables on the varying characteristics of these cases. Coded cross cultural research is a major field of study but has major limitations that need to be considered and overcome to deal with understanding the effects of common history and interaction among societies as part of the puzzle of scientific study of cultural causalities. Although I have been very involved at the evolution of this database and various experiments on how to study cultural and intercultural causation, I was for much of my career, until recently, quite annoyed with the tendency of those who contribute to cross-cultural analysis to ignore Galton's problem of modeling; namely, to study the interdependence among human societies and cultures (with the SCCS as sample cases). Some have felt that Galton's problem can easily be bypassed by random selection of cases for study, but the problem of historical connection or interaction is not so simply effaced. Probabiity samples are intended to insure independence of how samples are selected for study, but do not abolish the need to account for historical interdepencence of societies around the globe.

Because I and others have felt that cross-cultural comparisons are very important, although also problem ridden. Various attempts have been made recently to address this and other problems. Since 1969 on and off during my professional career,I have joined others in collaborative work to save valuable comparative data on the societies studied by anthropologists across many many decades in a vast majority of the world's geographic zones. Although in recent decades I have equally concentrated , modeling of macro and micro dat, on many other problems in sociology, history, networks and systems theory. Many of my concerns, much like Eric Wolf and Charles Tilly, has been to explore the kinds of impact produced by larger spatio-temporal processes (Tilly Big Structures, Large Processes, Huge Comparisons).I have worked on historical analysis, but also micro-ethnography carried out through the analysis of kinship networks and social processes, and a wide array of other ethnographic concerns on structure and system.

I returned most recently once again to consider finding solutions to valid cross-cultural analysis,. More specifically, stimulated by new kinds of solutions provided by a colleague in institutional economics and a former student I have recently returned to questions of how new modes of analysis can help to resolve not only Galton's problem and the dilemmas of cross-cultural research but to work with others to provide a new foundation for a new science of the study of causality in cultural systems.

Peter Medawar notes, in his older but very informative book, Advice to Young Scientists, that the basic question of science involves the quasi-experimental question, "I wonder what would happen if..." and the historical problems "How would I show what happens in...?" Regardless of what kind of empirical study a researcher does, the foundational questions are those that engage the experimental mode, which may differ according to the field one is in.

Cross-cultural and cross-national patterns matter. They involve and are implicated in the larger processes, such as uneven development and globalization, the rise and demise of cultural groups and polities. and affect how each of our societies and neighborhoods becomes embedded and interacts with others, as well as how a group transits to meet new challenges. A good example of this latter point is the ethnographic study begun in 1964 by Elizabeth Colson developed as the baseline for a "before and after" study of the effects of planned social change in the form of relation of Zambian villages prior to the constructing the Zambezi River Dam. (details, explain well and thoroughly.) A major basis of scientific activity as

## Sketch of the Causal graphs comparative ethnography project

### fyll and fydd: **Differentiation and Shame** in world cultures

- Effects of spatial clusters are always positive or insignificant if negative (as in a handful of cases).
- Effects of language phylogeny, however, when significant, are split between significant positive and negative effects, which is unexpected.
- 12 Results with negative fyll p >.10 that involve unusual behavior founders) such as warfare, violence, homicide, parental reaction to violence, wifebeating, police, money, and (the unusual feature) women's equal role as mythical founders.
- 19 Results with negative fyll p >.10 or positive fyll without these kinds of depvars
- no exceptions detected as yet, classifying rape in the latter category

- 11 1
- 2 17 correlation, Fisher exact p < .00001 (2-tailed)

- My interpretation is that negative effects of language phylogeny are am indicator either of differentiation (e.g., police, money, parental reaction to violence) or differentiation deviation from historical continuity for a behavior that is considered as "something you might do to your neighbors but would be shameful if done to your language mates or that they knew about" (warfare, violence, homicide, wifebeating). Rape, unfortunately, seems not to fall into this latter category by this criteria, judging from the nonsignificant effects of language phylogeny on its distribution.
**Significantly Negative fyll p <eq .10**- EduMod-28: (your dep var here) Imputation and Regression http://bit.ly/czBFiI v666 Interperviol
- EduMod-10: (your dep var here) Imputation and Regression http://bit.ly/d1SZic v666 Interperviol
- EduMod Mac-10: Imputation and Regression http://bit.ly/dmAR7p v666 Interperviol
- EduMod-15: (your dep var here) Imputation and Regression http://bit.ly/cU3lUz vv1772 Violence
- EduMod-16: (your dep var here) Imputation and Regression http://bit.ly/dhqIri v1746 React2viol
- EduMod-12: (your dep var here) Imputation and Regression http://bit.ly/b5RLr1 v1675 Homicide
- EduMod-18: (your dep var here) Imputation and Regression http://bit.ly/9jNRVZ v752 Wifebeating
- EduMod-22: (your dep var here) Imputation and Regression http://bit.ly/9UsACF v155 Money
- EduMod-59: Imputation and Regression http://bit.ly/cabt2k v155 Money
- EduMod-48: Imputation and Regression http://bit.ly/a5qDew v679 Warfight
- EduMod-56: Imputation and Regression http://bit.ly/9B2rFj v577 Mythical founders=Women ***
- EduMod-65: Imputation and Regression http://bit.ly/b03iNG v90 Police
- Negative nonsignificant fyll p >.10
- EduMod-17: (your dep var here) Imputation and Regression http://bit.ly/butHOz v122 Games of strategy
- EduMod-23: (your dep var here) Imputation and Regression http://bit.ly/czsN8t v26 Bodily contact
- EduMod-27: (your dep var here) Imputation and Regression http://bit.ly/9rhDaF v1721 Wealthy
- EduMod-66: Imputation and Regression http://bit.ly/9SSgni v1721 Wealthy
**Positive or nonsignificant fyll p >.10**

- EduMod-31: Imputation and Regression http://bit.ly/aHg1mF v892 Extwar *** exceptional: but there is reciprocal causation between v893 and v892, attacking and being attacked, and this occurs WITHIN as well as between language groups, which might indicate factional or segmentary war. This hypothesis is confirmed in the map fpr variable
**Continuous reciprocal external war**, defined below. These conflicts are not associated with v570 fratintgrps or any of the other measures of Fraternal Interest groups.

depvar<-apply(SCCS[,c("v892","v893")], 1,min) depvar<-(depvar==3)*1

- [[EduMod-11 667 rape Amanda (nonsignificant but compute whether fyll + or -)
- [[EduMod-41 667 rape - Christina (nonsignificant but compute whether fyll + or -)
- [[EduMod-13 678 foodstress Lawrence
- [[EduMod-21 678 famine Ralph
- [[EduMod-20 169 extramarital sex Bui
- [[EduMod-24 167 pre_mar_sex Michael
- [[EduMod-26 740 marr_arranged Abiha
- [[EduMod-29 591 Ownership or Control of the Use of Dwellings Kimberley
- [[EduMod-30 740 Marriage Arrangements=6 Parental control Jessica
- [[EduMod Mac-6 33 pain infliction Ryan (R2 near zero!) in ceremonials
- [[EduMod-47 901 loCasualties Doug
- [[EduMod-50 570 fratintgrp Doug
- [[EduMod-53 862 sororal polygyny Doug (R2 .34)
- [[EduMod-54 860 general polygyny Doug (R2 .42)
- [[EduMod-33 incomplete
- [[EduMod-51 incomplete
- 2010
- [[EduMod-55 661 "FemPartic" Chelsea
- [[EduMod-57 821 "PctFemContAg" User:Jasiellt
- [[EduMod-58 654 "spiritAggression" Bryan
- [[EduMod-64 678- "famine" Leon
- [[EduMod-67 1189 "EvilEye" Evan
- [[EduMod-70 1472 "FormalEd" Jeff

- Eliminate ambiguous results (different models for the same depvar)

- Negative fyll p >.10

None

- Negative fyll p <eq .10

none

### Causality is not functional consistency

see: Two Transitive Triples. at Indep/Depvar list - All

The Two Transitive Triples image (click) show causal estimate coefficients for regression effects (directed arrows from independent to dependent variables) that form transitive triples. Here it is the raw regression coefficients, positive (black) or negative (red) arrows, that are the quasi-experimental causal estimates. **Disregarding** the influences of clusters of neighboring societies and of common history is each language phylogeny, these coefficients measure the effect of increasing (or decreasing) this variable by one unit of measurement on the unit of measurement of the dependent variable, **holding all the other independent variables constant.**

The image shows evil eye beliers in SCCS societies, other independent variables held constant, are increased by milked animals and money (yellow and green nodes on red), while milked animals and money decrease the use of moned. See EduMod-27 and EduMod-59. This transitive triad has two positive links and one negative, an "inconsistent" triad but one consistent with causal effects. The **direct** effect of milked animals is 96 times stronger than the indirect effect.

### The causal coefficient

Note that the regression coefficients of 2SLS with social (Galton's) effects captured in first stage regression are **comparable** in that units of the independent variable have comparably scaled effects of the units of the shared dependent variables.

### Consistent node clustering and absence of node clustering

The colors of lines (red/black) attached to a node can be reversed when the polarity of the node is changed by reversing the direction of its categories: hi to low, low to hi. This *can* create sets of nodes with all-black links between them and all-red links to other clusters, *or vice versa*. There is no intrinsic reason for this to be the case, but network algorithms can determine a scale from 0 to 1 to which there is potentially consistent clustering (1) or maximal inconsistency (0), or some random mixture of the two in between.

Highly consistent clustering might be indicative of functional consistency, random inconsistency can indicate the lack of functional consistency. Maximum inconsistency as distinct from random inconsistency is not expected, just as negative distance effects are not expected.