Table 1.
Fit indices for latent class growth models, with the selected 4-class model indicated by shading.
| No. of classes |
BICa | SABIC | Log Likelihood |
Entropy | Average class probabilities | Adjusted LMRT b, c |
BLRT b,d |
|---|---|---|---|---|---|---|---|
| C1 | 7198.8 | 7182.9 | -3582.7 | -- | -- | -- | -- |
| C2 | 7270.0 | 7254.1 | -3618.3 | .87 | .96, .97 | --f | -- f |
| C3 | 7121.5 | 7096.1 | -3534.1 | .75 | .81, .94, .90 | < .001 | < .001 |
| C4 | 7076.4 | 7041.5 | -3501.5 | .73 | .92, .89, .73, .77 | < .001 | < .001 |
| C5 | 7082.8 | 7038.3 | -3494.7 | .71 | .82, .67, .90, .85, .60 | 0.1336 | < .001 |
| C6 | 7091.6 | 7037.7 | -3489.1 | .63 | .70, .90, .76, .73, .61, .59 | 0.2217 | E e |
Note:
a
Decrease greater than 6 indicates “strong” improvement in fit [52]
b
P-value, a lower p-value indicates that the k-class model is preferable over the k-1 class model.
c
Lo-Mendell-Rubin Adjusted likelihood Ratio Test
d
Parametric Bootstrapped Likelihood Ratio Test
e
Not all the bootstrap draws converged.
f
These two tests were not applicable, since within-class variances were specified differently across these two models.