开云体育Chuding, a couple of things. ? 1. In both articles, the unit of analysis is the dyad. So that measure of cohesion has a value for each pair of nodes. Is that what you want? The structural holes routine aggregates that cohesion to the node level (and yes, it is called indirects). So it depends on what the unit of analysis is in your study. ? 2. The correlation between indirects and constraint is hard to predict, as constraint is a function of not just ego network density but also network size and the degree distribution within the ego network. In general, though, indirects will be positively correlated with density and you would not want to subtract from 1. ? 3. For what it’s worth, I think the indirects (whether dyadic or monadic) is good for measuring constraint, but not ideal for cohesion. This is because it is a mixture of density and alter degree centralization (heterogeneous degree distribution). Centralization of alter degree almost always makes sense for constraint but only occasionally for cohesion. Why not use something directly interpretable, such as density? ? steve ? From: ucinet@...
Sent: Friday, November 23, 2018 23:34 To: ucinet@... Subject: [UCINET] How to calculate Reagan & McEvily’ (2003) social cohesion in UCINET? ? ? Dear Prof. Borgatti and colleagues, ? I am new to UCINET. Now I am trying to obtain the value of “social cohesion”, which is first introduced in Reagan & McEvily (2003). And later Tortoriello, Reagans, & McEvily (2012) give a more detailed explanation as follows: ? ? The way I tried to get this is using the routine in UCINET as follows: Ego Networks → Structural Holes. In the results, in addition to the traditional measures of structural holes mentioned in Borgatti et al. (2013/2018) like effect size, efficiency, constraint, and hierarchy, I also obtained a variable named as “indirects”. I am wondering whether this is the so called “indirect structural constraints” defined by Burt (1992). ? But after checking the correlations between these five SH variables, I found “indirects” was positively correlated to effect size while negatively correlated to constraint (raw value without transformation by 1 - constraint). So I guess if the “indirects” is the right variables I can utilize to obtain the measure of “social cohesion”, I need to do the transformation like “cohesion = 1 - indirects”. Am I right? Many thanks in advance! ? Best, Chuding |
