Table 6.

Descriptive statistics and correlations among constructs using machine learning (N=58,357 discussion threads).

Variablea	Mean (SD)	1	2	3	4	5	6
1. Positive emotional disclosure	1.27 (0.26)	1
2. Negative emotional disclosure	2.29 (1.08)	.19	1
3. Positive informational disclosure	1.67 (0.41)	.83	.14	1
4. Negative informational disclosure	3.69 (1.26)	–.05	.67	.04	1
5. Question	5.26 (1.29)	–.53	.16	–.48	.50	1
6. Provide emotional support	2.64 (1.06)	.24	.23	.24	.17	–.13	1
7. Provide informational support	2.95 (1.18)	–.22	.08	–.20	.21	.36	–.16

^a All variables were estimates from machine learning models predicting the amount a message contained the constructs in the first column, rated on a 7-point Likert scale where 1=not at all and 7=very much. The self-disclosure and question measures were based on the thread-starting message, whereas the measures of emotional and informational support were based on the first reply received in the thread.