Table 2.
Feature | Cohen’s d | p-value (corrected) | Mean diff-of-diff | 95% CI |
---|---|---|---|---|
Emergency | ||||
Change in Linguistic Style/ Mental Well-being | ||||
Categories that increase in usage before emergency visit | ||||
Anxious | 0.241 | <0.001 | 0.070 | [0.1, 0.38] |
Depressed | 0.238 | <0.001 | 0.066 | [0.1, 0.37] |
Arousal (how exciting the post is) | 0.191 | 0.011 | 0.051 | [0.06, 0.33] |
Valence (positive affect) | 0.168 | 0.017 | 0.079 | [0.03, 0.3] |
#posts b/w 9am-12pm | 0.160 | 0.020 | 0.016 | [0.02, 0.3] |
#posts b/w 12-3 pm | 0.144 | 0.047 | 0.013 | [0.01, 0.28] |
Categories that decrease in usage before emergency visit | ||||
Netspeak (‘u’, ‘da’, ‘smh’) | −0.374 | <0.001 | −0.01 | [−0.51, −0.24] |
1st person singular (‘I’, ‘my’, ‘me’, ‘I’m’) | −0.345 | <0.001 | −0.012 | [−0.48, −0.21] |
Informal Speech (‘lol’, ‘:)’, ‘b’) | −0.345 | <0.001 | −0.012 | [−0.48, −0.21] |
Leisure (‘family’, ‘fun’, ‘play’, ‘nap’) | −0.225 | 0.001 | −0.002 | [−0.36, −0.09] |
Change in Linguistic Topics | ||||
Topics that increase in usage before emergency visit | ||||
hospital, pain, surgery, blood, doctor, nurse | 0.230 | 0.001 | 0.001 | [0.09, 0.37] |
kids, child, their, children, mother, father | 0.165 | 0.021 | <0.001 | [0.03,0.3] |
even, still, tho, though, yet, blah | 0.165 | 0.021 | <0.001 | [0.03,0.3] |
thankful, very, amazing, most, blessed, wonderful | 0.142 | 0.013 | <0.001 | [0.01,0.28] |
Topics that decrease in usage before emergency visit | ||||
luv, nite, sum, 2day, kidz, doin | −0.315 | <0.001 | −0.001 | [−0.45, −0.18] |
<3, tht, lovin, bt, missin, ima | −0.308 | <0.001 | <0.001 | [−0.44, −0.17] |
nite, fb, bed, gn, sleep, night | −0.295 | <0.001 | −0.001 | [−0.43, −0.15] |
jus, bored, crib, house, chilin, hmu | −0.287 | <0.001 | −0.001 | [−0.42, −0.14] |
*Positive cohen’s d indicates an increase in the given style, while a negative score indicates decrease. Effect sizes of individual linguistic features (diff-of-diff b/w true and null events) for emergency visits. Significance was measured using paired, two-tailed t-test with Benjamini-Hochberg p-correction.