Skip to main content
. 2018 Jun 26;13(6):e0199286. doi: 10.1371/journal.pone.0199286

Table 3. Multinomial logistic regression results of predicting latent class membership, 2015(N = 2805) a,b.

Predictors High-risk group Moderate-risk group High-physical-inactivity and suicide-risk group
OR 95% CI OR CI (95%) OR 95% CI
Sex Girls - - - - - -
Boys 2.662 1.753–4.044 3.196 2.668–3.829 0.743 0.572–0.964
Grade 7th - - - - - -
10th 1.357 0.751–2.452 1.878 1.457–2.419 1.072 0.777–1.479
Monthly allowance Low - - - - - -
Medium 1.783 0.667–4.785 1.203 0.859–1.684 0.974 0.648–1.465
High 5.804 2.142–15.721 2.459 1.714–3.527 1.274 0.803–2.020
Parental relationship Good - - - - - -
Somewhat poor 2.161 1.309–3.567 1.504 1.183–1.912 2.266 1.680–3.058
Poor 4.245 2.347–7.679 1.251 0.852–1.838 2.992 1.973–4.537
Academic performance Good - - - - - -
Moderate 2.251 0.523–9.686 2.385 1.315–4.328 1.599 0.672–3.805
Poor 4.204 0.922–19.162 3.869 2.042–7.330 3.625 1.463–8.983

OR, odds ratio; CI, confidence interval; (-) not applicable and as the reference.

a Latent class analysis with 4 latent subgroups fit better than the models with 1, 2, 3, and 5 latent subgroups based on fit statistics (i.e., lower values for Akaike information criterion, Bayesian information criterion (BIC), and sample-size-adjusted BIC, higher values for average classification probability, and a low P-value for the Lo–Mendell–Rubin likelihood ratio test, indicating that the 3-class model had to be rejected in favor of a model with at least 4 classes).

b Table 3 presents the results of LCA multinomial regression analyses examining predictors of latent class membership with “the low-risk group” specified as the reference.