Table 4.

Results of the hurdle negative binomial regression models.

Variables		Model 1a (n=384,792b)		Model 2c (n=384,792b)		Model 3d (n=384,792b)		Without promotion (n=370,670b)		With promotion (n=14,122b)
		Coefficient	P value	Coefficient	P value	Coefficient	P value	Coefficient	P value	Coefficient	P value
Independent variables
	Beverage	0.05	.94	0.64	.36	–0.23	.12	–0.20	.18	–1.38	<.001
	Coffee	–1.69	<.001	–0.69	.003	–0.15	.21	–0.10	.40	–1.53	<.001
	Sweet	0.11	.496	–0.31	.06	–0.34	.001	–0.35	.002	–0.04	.81
	Dessert & bakery	–0.46	.09	–0.64	.04	–0.52	<.001	–0.57	<.001	–0.08	.78
	Fruits	–0.31	.004	–0.67	<.001	–0.24	<.001	–0.25	<.001	–0.07	.48
	Herbs & spices	0.20	.70	–1.19	<.001	–0.60	<.001	–0.61	<.001	0.14	.69
	Menthol & mint	–0.24	.34	–0.01	.97	–0.08	.46	–0.06	.58	–0.24	.55
	Nutty	–0.06	.86	0.01	.97	0.002	.99	–0.04	.79	0.65	.08
	Cream	0.16	.38	–0.39	.02	–0.15	.27	–0.19	.18	0.52	.007
	Tobacco	–1.65	<.001	–0.94	.06	–0.55	.001	–0.54	.001	–0.66	.10
	Chocolate	–0.14	.63	0.28	.57	0.09	.65	0.10	.62	0.13	.59
Control variables
	Promotion			3.42	<.001	3.16	<.001
	Posts (Coefficient x 10–4)			—e	—	3.58	<.001	3.59	<.001	2.10	<.001
	Average share (Coefficient x 10–2)			—	—	10.17	<.001	.11	<.001	3.20	<.001
	Topic 1			—	—	1.39	<.001	1.19	<.001	3.54	<.001
	Topic 2			—	—	1.17	<.001	0.73	.008	10.26	<.001
	URL mention			—	—	0.19	<.001	0.20	<.001	0.01	.89
	Hashtag			—	—	–0.02	.72	0.005	.94	–0.12	.24
Intercept		–18.18	<.001	–16.96	<.001	–18.79	<.001	–15.38	<.001	–2.42	<.001

^aEstimates of a specification of the initial model with only flavors as the independent variables.

^bNumber of observations is defined by n.

^cEstimates of flavors’ influence on information propagation modified by adding promotion only.

^dEstimates of flavors’ influence on information propagation modified by additional control variables.

^eVariable not used in the second "All data" model.