# Using personality to predict conflict via ERGMs

I am using ERGMs to test how different personality variables lead to conflict connections among group members. Specifically, I am assessing if differences in extraversion (Big Five) among team-member dyads make it more likely for conflict to occur. I have been using the absdiff function to test this as shown here:

c<-ergm(RC3MS~edgecov(structural0)+edges+absdiff("Extra")+nodecov("English")+nodefactor("Country"))


In this case "Extra" refers to extraversion, and the other parameters are control variables among some other model specifications. The issue I am having is that the absdiff function can only test if overall conflict is more likely to emerge in these pairs without taking into account the direction of the conflict (i.e., who is reporting conflict within the dyad). Does anyone know of a function that could assess this relationship, while also considering the direction of the reported conflict? Thank you!

Check out the nodeicov() and nodeocov() terms. I'm assuming (can't comment yet!) that your edges represent conflict between ties, where $i \rightarrow j$ indicates that $i$ reports a conflict with $j$. The first term would tell you if individuals with greater extraversion scores are more likely to have in-directed conflict ties. The second would tell you if greater extraversion is associated with more out-directed conflict ties.
Another option: Make "Extra" categorical (e.g. "high", "medium", "low") and use the categorical homophily and assortive mixing terms (nodematch() [use the argument diff=T if you expect differential homophily], nodemix()). This will let you explore the likelihood of tie formation among pairs of nodes in each category.
• I will add that the nodeicov and nodeocov are not what I am looking for here, as I am looking for the interaction among dyads and not singular effects. Also extraversion is a continuous personality variable, so I would prefer not to categorize it. Additionally I am postulating that as the magnitude of the difference between these variables increases conflict is likely to occur. Thus, if I categorize it I lose essentially all meaning of magnitude as one cannot technically assess this with categories. So to recap, I am purely looking for the direction of the conflict within these dyads. – costebk08 Oct 24 '16 at 4:58
• I see what you mean. In that case, the edgecov and dyadcov terms both accept matrices of dyadic covariates. You could create an n x n matrix of the difference extraversion scores. A negative score indicates that the reporter of the conflict is less extroverted than the person they report (i.e. in $i \rightarrow j$, $i$ is more introverted). A positive score indicates that the reporter is more extroverted. The coefficient on the term would indicate whether there is a relationship between the difference in extraversion between any given dyad and the directionality of that relationship. – paqmo Oct 24 '16 at 15:25
• As you mentioned those functions only take matrices of dyadic covariates, and the extraversion data is continuous; thus, it will not work. I am basically looking for a function that does the same thing as absdiff but with directionality. I really do appreciate that effort though thank you! – costebk08 Oct 27 '16 at 15:23
• Just to clarify, dyadic covariates can indeed take continuous terms. Dyadic here indicates that the covariate is the property of a dyadic relationship, not that it is binary. In this case, make a matrix of the differences in extraversion score for each $i, j$ pair. Looking across each row, the negative entries indicate the sender is less extroverted than the receiver, and vice versa. This, therefore, does the same thing as absdiff, but admits directionality through the sign of the each entry. – paqmo Oct 27 '16 at 16:32