Table 3 (Left) Correlation of our language categories with behavioral markers computed with alternative techniques and datasets.

From: How epidemic psychology works on Twitter: evolution of responses to the COVID-19 pandemic in the U.S.

     

Correlation with phases

 

Marker

ost correlated language categories

Refusal

Anger

Acceptance

Custom words

Alcohol

body (0.70)

feel (0.62)

home (0.58)

−0.43

0.46

−0.12

 

Economic

anxiety (0.73)

negemo (0.68)

negate (0.56)

−0.12

0.37

−0.53

 

Exercising

affiliation (0.95)

posemo (0.93)

we (0.92)

−0.62

0.31

0.89

Interactions

Conflict

anxiety (0.88)

death (0.57)

negemo (0.54)

0.58

−0.24

−0.92

 

Support

affiliation (0.98)

posemo (0.96)

we (0.94)

−0.68

0.37

0.90

 

Power

prosocial (0.95)

care (0.94)

authority (0.94)

−0.48

0.18

0.88

Medical

Physical health

swear (0.83)

feel (0.77)

negate (0.67)

−0.66

0.81

−0.32

 

Mental health

affiliation (0.91)

we (0.88)

posemo (0.85)

−0.65

0.36

0.85

Mobility

Travel

death (0.59)

anxiety (0.58)

 

0.62

−0.32

−0.82

 

Grocery

I (0.80)

leisure (0.72)

home (0.64)

−0.77

0.70

0.29

 

Outdoors

sad (0.68)

posemo (0.65)

affiliation (0.59)

−0.62

0.39

0.72

  1. For each marker, the three categories with strongest correlations are reported, together with their Pearson correlation values in parenthesis. (Right) Pearson correlation between values for our behavioral markers and “being” in a given phase or not. Values in bold indicate the highest values for each marker across the three phases. All reported correlations are statistically significant (p < 0.01).