At the same time, trying to save
different types of information about whole network, we incorporate information
on non-respondents into models by following ways[12].
First, as a provisional trade-off, we will create attribute variables that
capture some information on ties to non-respondents. These will be, for
example, “ties-to-non-respondents” variable, mere counter of number of such
ties. We also will try to reflect known structural features of non-respondents,
and weight ties to them according with some of their centrality indexes, such
as centrality, or in-degree neighborhoods with various decay factors.