Table 4.
A comparison of model fits: models with indicators of HIV infection and transmission only (model 1) versus models with Facebook features added (model 2).
| Dependent variable and model | LRa chi-square (df) | F test (df) | Model 1 vs Model 2 LR chi-square (df) | Model 1 vs Model 2 F change (df) | P value | |
| Condomless sex |
|
N/Ac | 18.2 (4)d |
|
.001 | |
|
|
Model 1b | 6.8 (1) |
|
|
|
|
|
|
Model 2e | 25.0 (5) |
|
|
|
|
| Sex drug use |
|
N/A | 21.4 (5)d |
|
<.001 | |
|
|
Model 1 | 16.5 (1) |
|
|
|
|
|
|
Model 2 | 37.9 (6) |
|
|
|
|
| Biomedical prevention |
|
N/A | 11.5 (2)d |
|
.003 | |
|
|
Model 1 | 31.6 (2) |
|
|
|
|
|
|
Model 2 | 43.2 (4) |
|
|
|
|
| Depressionf | N/A |
|
N/A | 3.14 (6, 302)d | .005 | |
|
|
Model 1 |
|
16.88 (1, 308) |
|
|
|
|
|
Model 2 |
|
5.21 (7, 302) |
|
|
|
aLR: likelihood ratio.
bFor each outcome, model 1 includes all HIV infection and transmission risk indicators that met a Cronbach α=.10 criterion in previous models.
cN/A: not applicable.
dThe statistical significance of the comparison of the two models.
eFor each outcome, model 2 adds to model 1 the Facebook network and communication variables that met a Cronbach α=.10 criterion in previous models.
fDepression is a numeric measure, so an F test and F change are reported instead of LR chi-square tests.