Table 3.

Mathematical and statnet specifications of tested network formation models.

Model	Mathematical specification	Statnet specification
Null	Pr(X = sources_net|θ)= $\frac{1}{\mathrm{\kappa }(\mathrm{\theta })}($θ0*edges)	M0 ← ergm(sources_net ~ edges)
Node Attributes Only	Pr(X = sources_net|θ)=$\frac{1}{\mathrm{\kappa }(\mathrm{\theta })}($θ0*edges+θ1*absdiff('Toxicity')+θ2*nodecov('Toxicity')+θ3*absdiff('Upvote_Ratio')+θ4*nodecov('Upvote_Ratio')+θ5*absdiff('Number_of_Comments')+θ6*nodecov('Number_of_Comments')+θ7*absdiff('Word_Count')+θ8*nodecov('Word_Count')))	M1 ← ergm(sources_net ~ edges+absdiff('Toxicity') +nodecov('Toxicity') +absdiff('Upvote_Ratio') +nodecov('Upvote_Ratio')+ absdiff('Number_of_Comments')+nodecov('Number_of_Comments')+absdiff('Word_Count')+nodecov('Word_Count'))
Structural and Node Attributes Model	Pr(X = sources_net|θ)=$\frac{1}{\mathrm{\kappa }(\mathrm{\theta })}($θ0*edges+θ1*absdiff('Toxicity')+θ2*nodecov('Toxicity')+θ3*absdiff('Upvote_Ratio')+θ4*nodecov('Upvote_Ratio')+θ5*absdiff('Number_of_Comments')+θ6*nodecov('Number_of_Comments')+θ7*absdiff('Word_Count')+θ8*nodecov('Word_Count') +θ9*gwesp(0.2, T)))	M2 ← ergm(sources_net.filt ~ edges +gwesp(0.2, T) + absdiff('Toxicity') +nodecov('Toxicity') +absdiff('Upvote_Ratio') +nodecov('Upvote_Ratio') +absdiff('Number_of_Comments') +nodecov('Number_of_Comments'))