-
Notifications
You must be signed in to change notification settings - Fork 0
Description
From @TomSnijders:
In the new version of RSiena, sienaGOF objects have an attribute EgoAlter, which is TRUE currently only for egoAlterCombi.
As we discussed earlier, the purpose is to signal to plotting routines that the results have an ego-by-alter structure.
The plotKey (given as argument 'key' to test_gof, and by default attr(x, "key") where x is the sienaGOF object) for the ego-by-alter structure is ppnames defined by
- brange is the range of the monadic variable.
- trafo is an increasing function, should have values from 0 to 9 or a smaller range
- possible.pairs <- sort(unique(as.vector(outer(10*trafo(brange), trafo(brange), '+'))))
- ppnames <- as.character(possible.pairs)
The help page says "The range of this transformation should preferably have 5 or fewer values, and definitely not 10 or more, as this would lead to problems in the coding of ego-alter combinations".
ppnames is the plotKey and the vector of names of values of the auxiliary statistic.
In words, the plotKey is a number composed of two digits; the first digit refers to the value of ego and the second to the value of alter:
ego = plotKey %/% 10
alter = plotKey %% 10
egoAlterCombi is specifically for dependent behavior variables. I plan to construct another auxiliary statistic that does just the same thing for arbitrary monadic variables (covariates or other).
It would be great if autograph could make a plot exploiting this ego-alter structure!
