search.ergmTerms behavior -- returns bipartite terms for non-bipartite network about ergm HOT 6 OPEN

This is related to #409, which we never resolved. Briefly, we can have keyword "bipartite" be set on terms that work on bipartite networks (including, e.g., absdiff) or on terms that only work on bipartite networks (e.g., b1cov). Carter and I had a back-and-forth, but there wasn't any resolution that was implemented.

from ergm.

I wish the world was simpler. That said:

1. I would vote in favor of the former -- I think most people will use the keyword to find terms that can be used in their networks, not the terms that can ONLY be used in their networks.

2. We still shouldn't be returning bipartite only terms for non-bipartite nets.

from ergm.

I wish the world was simpler. That said:

1. I would vote in favor of the former -- I think most people will use the keyword to find terms that can be used in their networks, not the terms that can ONLY be used in their networks.

2. We still shouldn't be returning bipartite only terms for non-bipartite nets.

And therein lies the problem: you can have one or the other but not both.

Let's say we go with the "works on bipartite". Then, both absdiff and b1cov would have the keyword, so both will show up for a bipartite network; on the other hand, you then won't be able to use it to filter out b1cov on faux.mesa.high. So, we can do 1. but not 2..

from ergm.

Aha, so the problem is that we don't have a keyword for basic (unipartite, undirected, non-hyper, non-multiple, binary) nets like faux.mesa.high?

Is it possible to modify this code to exclude terms that start with "b" or "d" if the nw attributes are all "FALSE" for those conditions?

As a side note -- I'm not sure the keywords have been updated for "multiple". If so, the documentation hasn't.

from ergm.

Aha, so the problem is that we don't have a keyword for basic (unipartite, undirected, non-hyper, non-multiple, binary) nets like faux.mesa.high?

That could be a possible remedy. "Unipartite" is a very ugly word, but adding it to all terms but the explicitly bipartite ones should, in principle, solve the problem.

Is it possible to modify this code to exclude terms that start with "b" or "d" if the nw attributes are all "FALSE" for those conditions?

It's possible, but it would be a very bad idea to make assumptions about what a term does based on its name. A more reliable approach might be to try to parse the body of the Init*ErgmTerm.*() function and the check.ErgmTerm() call, I suppose, but that also assumes that the term author writes a term in a particular way.

As a side note -- I'm not sure the keywords have been updated for "multiple". If so, the documentation hasn't.

I am not sure what you mean. There is no "multiple" keywords; are you referring to binary vs. valued?

from ergm.

I am not sure what you mean. There is no "multiple" keywords; are you referring to binary vs. valued?

no -- to multinet modeling terms, if there are any

from ergm.