How to compute whether bicomponent cohesion increases or decreases with generation in Pajek kinship data

From InterSciWiki

Jump to: navigation, search

The kinship network should be in Pajek p-graph format, couples as nodes.

Contents

[edit] Object

The OBJECT here is complicated: to extract the connected component, then eliminate 1 or more bottom generation (unmarried children), then compute bicomponents, recompute generations, then use /Partitions to correlate depth of generation by percent of marriages in bicomponent (as we did for Omaha) -- is cohesion increasing, decreasing or stable?

[edit] Remove lower generations

Lower generations are usually thin because couples don't have their children yet/

/Net/Partition/Depth/Generations
/Draw/Draw-partitions
/Operations/Extract from network/Partitions and enter for example 3-* to exclude lower generations 1 and 2. Your network is now cleared of the two younger generations/
Again /Net/Partition/Depth/Generations

[edit] Take only the connected component

/New/Component/weak -- puts the components into different partitions. The /Operations/Extract/Partition (and give the number of the partition with the most people). How do you know? /Info/Partitions

[edit] Graph to get the generations

In Draw window /Layers/In y direction /Layers/Optimize layers in x direction/Forward

Spin in the z dimension to invert the graph so ancestors are at the top

[edit] Compute the bicomponent and make it a partition

  • Do /Net/Components/Bicomponent (click Hierarchy window -- THEN CLICK ROOT, you will get a list if bicomponents -- the sizes of the bicomponents will differ, jot down the number of the giant bicomponent, i.e., the largest)
i.e., Go to Hierarchy window after /Net/Component/Bicomponent

double click in the white space on the right of the Hierarchy window in the rectangle that opens there will be "Root" Click "Root" for the list of bicomponents and sizes.

  • Now do /Hierarchy/Extract cluster -- give the number (integer from series 1,2,... as listed) of the giant bicomponent
Now do /Cluster/Make partition -- enter the SIZE of the network as the SIZE of the partition. That is, when you convert your bicomponent to a partition you have to give it the size of the network.

[edit] Arrange the two partitions in the two different partition windows

Now you have your bicomponent partition, consisting of 0 and 1 for nodes in the bicomponent

Click on the right side the select icon for the SECOND partition WINDOW and click on the Generations partition

[edit] Compare the two partitions

  • You should now have the Bicomponent partition in the FIRST partition WINDOW and the Generations partition in the SECOND
/Partitions/Info/Spearman's rank
  • If Spearman's correlation is negative and less than -0.2 then the LOWER GENERATIONS are more cohesive; if positive and greater than 0.2 then the OLDER GENERATIONS are more cohesive.
The closer the correlation to -1 or +1 the more bicomponent cohesion is changing with time.

[edit] Draw your bicomponent partition

Assuming your bicomponent partition is in the FIRST partition WINDOW

/Draw/Draw-partitions -- now the Draw window should have ordered generations that have yellow nodes for the bicomponent and cyan nodes for nodes that lack structural endogamy.
The nodes should be in successive GENERATIONS - your result should be visible.
Retrieved from "http://intersci.ss.uci.edu/wiki/index.php/How_to_compute_whether_bicomponent_cohesion_increases_or_decreases_with_generation_in_Pajek_kinship_data"