Topic 1: Find_in_OreGraph.doc - D. White 13/6/2003
Here is the procedure for using pajek to find types of blood marriage or
relinking in a genealogical network in which the nodes are individuals.
The idea here is that you have a genealogical network (Ore graph) in
pajek format with triangles for males and ellipses for females, and you
click Net/Partitions/Vertex shapes to create a partition for gender of
individuals. Let's say that you want to compare that network and its
partition with a series of small networks each of which represents a
type of marriage.
To set up the comparison (with network and partition in their respective
windows) click Nets/Second Network and then click Partitions/Second Partition.
Now read in the smaller network you want to find within the larger one, click
Net/Partitions/Vertex shapes to get its gender partition, then click
Nets/First Network and click Partitions/First Partition.
Then click Nets/Fragment (1 in 2)/Options and check [x] labeled and [x]
extract subnetwork, making sure nothing else is checked. Then click
Nets/Fragment (1 in 2)/Find to find the subnetworks and report their
frequency.
I used the following subfiles to count frequencies of MZD, MBD, FZD
and FBD (considering only paternal siblings in the middle generation).
You can cut and paste them into files named MZD.net and so forth and
test them out on your data. Note in each case that the first node is a
triangle (male) and will be partition 1 in the Partition generated by
Net/Partitions/Vertex shapes so make sure that the same partition holds
for your larger genealogical *.net file.
Message to Houseman and our Working group (T.I.M.E.): it is now only
necessary to create similar files (by a program from Klaus's or
Laurent's formulae) for each type of blood marriage or relinking that we
wish to study.
MZD - The x,y,z coordinates of the nodes are not essential to any of
these files. Other files may be written without them. In the present
case they simply help to visualize the four graphs.
*Vertices 5
1 "" 0.6200 0.6200 0.5000 triangle
2 "" 0.6200 0.3800 0.5000 ellipse
3 "" 0.4840 0.1420 0.5000 triangle
4 "" 0.3800 0.6200 0.5000 ellipse
5 "" 0.3800 0.3800 0.5000 ellipse
*Arcs
3 2 1
2 1 2
3 5 1
5 4 2
*Edges
1 4 0
MBD
*Vertices 5
1 "" 0.6200 0.6200 0.5000 triangle
2 "" 0.6200 0.3800 0.5000 ellipse
3 "" 0.4840 0.1420 0.5000 triangle
4 "" 0.3800 0.6200 0.5000 ellipse
5 "" 0.3800 0.3800 0.5000 triangle
*Arcs
3 2 1
2 1 2
3 5 1
5 4 1
*Edges
1 4 0
FZD
*Vertices 5
1 "" 0.3800 0.6200 0.5000 triangle
2 "" 0.3800 0.3800 0.5000 triangle
3 "" 0.4840 0.1420 0.5000 triangle
4 "" 0.6200 0.6200 0.5000 ellipse
5 "" 0.6200 0.3800 0.5000 ellipse
*Arcs
3 2 1
2 1 1
3 5 1
5 4 2
*Edges
1 4 0
FBD
*Vertices 5
1 "" 0.3800 0.6200 0.5000 triangle
2 "" 0.3800 0.3800 0.5000 triangle
3 "" 0.4840 0.1420 0.5000 triangle
4 "" 0.6200 0.6200 0.5000 ellipse
5 "" 0.6200 0.3800 0.5000 triangle
*Arcs
3 2 1
2 1 1
3 5 1
5 4 1
*Edges
1 4 0
Andrej -
What I would like is to be able to define a "distinguished" node in each
of my subgraphs. For example, one of the top ancestors. And then each time
the fragment is found, to record which is the distinguished node, and to
increment the value of the partition for that node by one every time
a distinguished node occurs within a fragment. I can then cross this with
generational levels to get a frequency distribution for different time
periods. - Doug
back
(This is an expansion)
1. If you start from reading a GEDcom file in Pajek as an Ore graph (individuals as nodes)
you will have to set the Options/ReadWrite Threshold to a negative value (e.g. -1),
because you have 0 values of lines for edges representing marriages.
Otherwise edges (the marriage relation) will not be read.
If you want to take gender into account you must also set in Options/ReadWrite
Ore 1-male, 2-female. Because we use the gender of nodes in the main instructions above
it is not necessary in Nets/Fragments (1 in 2)/options to check values of lines as below.
Go directly to the first set of instructions above.
2. In Nets/Fragment 1 in 2/ Options
uncheck induced, since half siblings can have several children.
check values of lines if you want to take gender into account
3. Now to find a type of subgraph (relinking or blood marriage),
make a small *.net file such as this one which finds FBD marriage.
*Vertices 5
1 ""
2 ""
3 ""
4 ""
5 ""
*Arcs
3 2 1
2 1 1
3 5 1
5 4 1
*Edges
1 4 0
4. With this smaller graph in the network window,
Click Nets/First Network
With the total graph in the network window,
Click Nets/Second Network
then
Click Nets/Fragment (1 in 2)
5. The analogous graphs for MBD and FZD, however, or any case where there sex of the
couple needs to be define require an additional step to specify gender properly.
(because the value 2 of lines is the matriline whereas values of 2 for links to daughters
are needed to specify gender through line values alone).
6. In this case you must use a partition to define matching in gender - it is a little complicated:
7. For data file (*.net) compute
Net/Partitions/Vertex shapes [this gives a partition on the gender of individuals as nodes]
In the picture you will get only yellow and green vertices (partition values 1 and 2).
Define the obtained partition as second partition (Partitions/Second).
See attached 2.clu.
8. In your pattern network (subgraph.net) you must 'color' vertices in the same way
triangles must be of the same color as in the previously obtained network.
Define the obtained partition as first partition (Partitions/First).
9. In Fragments/Options check 'Labeled' - which means numbers (colors) in the partition
in the pattern that is to be found must also match those in the original network.
= == == == == == == == == == == == == == == == == == == == == =