net_x_triad() now takes a mixed census where a multiplex network holds
exactly one one-mode layer and one two-mode layer, after Hollway et al. (2017).
It falls through to the ordinary flat census in every other multiplex case.
That leaves an open question: what should it do with a multiplex network of
two or more one-mode layers, such as ison_algebra (social, tasks,
friends) or ison_monks (like, esteem, influence, praise)?
Today it flattens them and takes one Davis-Leinhardt census over the union of
the layers. That is defensible but discards the multiplexity, which is at odds
with the principle that the census should use as much information as the
network gives it.
A multiplex triad census over two one-mode layers is a different object from
the mixed census, and needs its own decision:
- which motifs it enumerates, and how they are labelled;
- whether it generalises beyond two layers;
- whether
node_x_triad() should report it per node, which it would do for
free through its existing leave-one-out construction at
R/motif_census.R:169.
net_x_triad()now takes a mixed census where a multiplex network holdsexactly one one-mode layer and one two-mode layer, after Hollway et al. (2017).
It falls through to the ordinary flat census in every other multiplex case.
That leaves an open question: what should it do with a multiplex network of
two or more one-mode layers, such as
ison_algebra(social, tasks,friends) or
ison_monks(like, esteem, influence, praise)?Today it flattens them and takes one Davis-Leinhardt census over the union of
the layers. That is defensible but discards the multiplexity, which is at odds
with the principle that the census should use as much information as the
network gives it.
A multiplex triad census over two one-mode layers is a different object from
the mixed census, and needs its own decision:
node_x_triad()should report it per node, which it would do forfree through its existing leave-one-out construction at
R/motif_census.R:169.