Table 2. Partial correlations between atherosclerotic heart disease (AHD) mortality and Twitter language measured by dictionaries, across the northern and southern halves of the United States.

	North			South
Language variable	Partial r	p	95% CI	Partial r	p	95% CI
Risk factors
Anger	0.240	<.001	[0.168, 0.310]	−0.020	.604	[−0.095, 0.056]
Negative relationships	0.156	<.001	[0.081, 0.229]	0.060	.121	[−0.016, 0.135]
Negative emotions	0.108	.005	[0.032, 0.182]	0.028	.462	[−0.047, 0.104]
Disengagement	0.166	<.001	[0.092, 0.239]	−0.012	.750	[−0.088, 0.063]
Anxiety	0.017	.654	[−0.058, 0.093]	0.104	.007	[0.028, 0.178]
Protective factors
Positive relationshipsa	−0.032	.411	[−0.107, 0.044]	0.111	.004	[0.035, 0.185]
Positive emotions	−0.166	<.001	[−0.238, −0.091]	0.082	.034	[0.006, 0.156]
Positive engagement	−0.192	<.001	[−0.264, −0.119]	0.041	.288	[−0.035, 0.116]

Notes.

Partial r: partial correlation coefficients obtained from a regression predicting AHD from the Twitter theme represented by the language variable, with county-level education and income as control variables.

^aFollowing Eichstaedt et al. (2015a) the word love was removed from the dictionary for this variable. See discussion in the text.