Table 2.
Linear growth models for the current BMI of 1170 users of an eating disorder community on Reddit.
|
|
Model 1 | Model 2 | Model 3 | Model 4 | |
| Fixed effect estimates (SE) | |||||
|
|
Intercept | −0.473 (0.025) | −0.165 (0.013) | −0.169 (0.013) | −0.165 (0.013) |
|
|
Time | —a | −0.074 (0.004) | −0.072 (0.004) | −0.087 (0.005) |
|
|
Time×baseline current BMI interactionb | — | — | −0.044 (0.004) | −0.043 (0.004) |
|
|
Time×activity interactionb | — | — | — | −0.118 (0.020) |
| Random effects variance | |||||
|
|
Intercept | 0.512 | 0.068 | 0.068 | 0.069 |
|
|
Residual | 0.829 | 0.250 | 0.251 | 0.250 |
|
|
Slope | — | 0.011 | 0.009 | 0.009 |
| Goodness-of-fit measures | |||||
|
|
Akaike information criterion | 15,172.918 | 10,504.541 | 10,387.654 | 10,353.948 |
|
|
Bayesian information criterion | 15,192.584 | 10,543.871 | 10,433.540 | 10,406.389 |
|
|
Log likelihood | −7583.459 | −5246.271 | −5186.827 | −5168.974 |
a—: not applicable.
bMain effects of baseline current BMI and activity were excluded from their models as their inclusion lowered the model fit.