Ac06 Ndembu

From InterSciWiki
Jump to: navigation, search

Ac06 Ndembu 1930

Turner, Victor. 1957. Schism and Continuity in an African Society: A Study of Ndembu Village Life. Manchester: The University Press.

Werbner, Richard P. 1984. The Manchester School in South-Central Africa. Annual Review of Anthropology 13:157-185,

A biography of Victor W. Turner

Kinship structure of the Ndembu: Appendix III, coded by Tolga Oztan

Victor Turner Flyleaf III A1-Nyachima Apical Ancestor of Matrilineage

Pajek Generational Partition Corrected: Turner Flyleaf III Tolga Oztan and Douglas R. White. Click to enlarge. THE GREEN NODE 4 GENERATIONS DOWN IS MISPLACED BY PAJEK AND PLACED AT THE UPPER GENERATION IN THE PAJEK SLIDE BELOW. Pseudocode at P-graph_generation_levels takes care of this problem, as shown in the current graph. The solution is to place the white node and ones like it just above their child, in the main DAG, and repeat this operation elsewhere.
Pajek Genealogical Partition: Showing the problem with Pajek's "Generation" Algorithm

The arrows are to parental nodes from nodes for their children, either single or parents themselves. The dotted red lines represent female links to parents and solid black lines are for male links to parents. This is called pgraph format which has only arcs and no directed cycles, hence forming a directed asymmetric graph which has a determinate number of separable generations but the placement of couples is not determinate. The objective of these graphs are to minimize the sum of distances in level for each parent-child pair, the minimum sum equaling the number of arcs. When two arcs converge to form a couple but their distance to parents differs, it may indicate that one spouse is older than the other.

The upper graph is a correct alignment of generations, starting from the apex of the longest number of vertical generations, and working downward, although some of the colors that should differentiate generations are askew. Once all those descendants are placed, the error in the Pajek net/partition/generation algorithm that colored the graph occurs in working up to another apical ancestor, erroneously colored as if it were as high as the highest ancestor. The algorithm should be changed so as to work from one of the previously placed descendants to a parent that is not among them, then up to another apical ancestor, not necessarily as high as the first. From each new apical ancestor it should then work down again to place descendants, as before.

To be successful, the sum of distances in level for each parent-child pair must be minimized. The actual position of the nodes achieves this with a score of 6 true generational differences between spouses but it is not always clear where the parental nodes for these couples should be located without altering the score of 6. The first algorithm has these 6 intrinsic errors plus 5 other created by the algorithm (colors out of place). The second algorithm has 6 + 12 of mistaken color placement. Why these occur and how they could be fixed is unclear.

The programmer cannot assume that all descent lines will end in the same lower generation or that all ancestor begin at the top generation.

Taking the transpose of the downward arrows to define a parent-to-children's node graph does not improve the results.