SciComm Optimizer for Policy Engagement: a randomized controlled trial of the SCOPE model on state legislators’ research use in public discourse

Table 2 Intervention effect modeled with zero-inflated negative binomial regressions

	Estimate (95% CI) p-value
	IRR (count model)			OR (logistic/inflation model)
Predictor	Intervention			Intervention				State professionalism (Squire Index)
Main effect (overall URE)	1.24	(1.04, 1.48)	0.02	1.15	(0.79, 1.68)	0.46	1.09		(1.08, 1.11)	0.00
Post hoc analyses
Accountability	17.03	(1.78, 162.82)	0.02	122.73	(1.98, 7631.20)	0.02	1.07		(1.03, 1.11)	0.00
Conceptual	1.28	(1.05, 1.56)	0.01	1.09	(0.74, 1.60)	0.68	1.09		(1.08, 1.11)	0.00
Data/analytics	1.67	(1.03, 2.07)	0.04	0.79	(0.23, 2.43)	0.64	1.11		(1.05, 1.18)	0.00
Knowledge generation	0.31	(0.15, 0.64)	0.002	0.32	(0.12, 0.92)	0.04	1.04		(1.01, 1.08)	0.01
Methods	0.56	(0.26, 1.19)	0.13	0.53	(0.21, 1.32)	0.17	1.04		(1.01, 1.07)	0.01
Problem definition	1.24	(0.09, 1.74)	0.20	1.70	(0.64, 4.51)	0.28	1.08		(1.04, 1.12)	0.00
Problem solution	1.22	(0.60, 2.47)	0.59	1.00	(0.38, 2.66)	1.00	1.04		(1.02, 1.06)	0.00
Type of study	11.37	(1.70, 76.08)	0.01	78.26	(0.57, 13,083.41)	0.10	1.04		(1.01, 1.08)	0.02

This is a zero-inflated model, such that the model simultaneously conducts a logistic regression to determine predictors of whether the legislator created any posts with URE (0 vs 1 + ; i.e., the zero inflation), and a negative binomial regression to determine predictors of how frequently the legislator created posts with URE among those with more than 0 posts (the count model). We modeled the inflation with the index of state professionalism and their experimental condition (Intervention), and the count with just the experimental condition. The columns under “Logistic/inflation model” report the OR (odds ratio) and associated numbers for those predictors, and the columns under “Count model” report the IRR (incident rate ratio) for the intervention indicator
Overall URE is the count of social media posts that had any of the following eight URE type categories that follow. Control group was the reference in these analyses

ISSN: 1748-5908