layers and edge covariates

Dear list,

in the example on Edge layers and covariates
<https://graph-tool.skewed.de/static/doc/demos/inference/inference.html#edge-layers-and-covariates&gt;,
blocks are fitted as follows:

state = gt.minimize_blockmodel_dl(g, deg_corr=False, layers=True,
                                  state_args=dict(ec=g.ep.value,
layers=False))

I'm trying to make sure I understand the LayeredBlockState correctly. Are
the following statements correct?

   1. The independent layers version is used, which means that there is one
   layer for every possible number of co-appearances. *This means that
   number of co-appearances is treated as a categorical, rather than an
   ordinal variable. *
   2. If one wanted to encourage the model to assort actors into the same
   block if they have many co-appearances, the following fit would be more
   appropriate:

   state = gt.minimize_blockmodel_dl(g, deg_corr=False, layers=False,
                                     state_args=dict(eweight=g.ep.value))

(If I'm right, then I find that the second model is closer to what an
applied scientist would be interested in...)

Many thanks for clearing this up,

Peter

attachment.html (2.68 KB)

Both statements are correct.

Best,
Tiago

Thank you Tiago
Peter

Dear list,

in the example on Edge layers and covariates
<

https://graph-tool.skewed.de/static/doc/demos/inference/inference.html#edge-layers-and-covariates

,
blocks are fitted as follows:

state = gt.minimize_blockmodel_dl(g, deg_corr=False, layers=True,
                                  state_args=dict(ec=g.ep.value,

layers=False))

I'm trying to make sure I understand the LayeredBlockState correctly. Are
the following statements correct?

1. The independent layers version is used, which means that there is one
    layer for every possible number of co-appearances. /This means that
    number of co-appearances is treated as a categorical, rather than an
    ordinal variable. /
2. If one wanted to encourage the model to assort actors into the same
    block if they have many co-appearances, the following fit would be

more

    appropriate:

    state = gt.minimize_blockmodel_dl(g, deg_corr=False, layers=False,
                                      state_args=dict(eweight=g.ep.value))

Both statements are correct.

Best,
Tiago

attachment.html (3.2 KB)