Skip to main content
. Author manuscript; available in PMC: 2022 Apr 1.
Published in final edited form as: J Fam Psychol. 2021 Apr;35(3):276–287. doi: 10.1037/fam0000646

Table 2.

Results for Preliminary LPA Models

Model 1: Class-specific means with class-invariant variances, covariances between correlated indicators Model 4: Class-specific means, variances and covariances between correlated indicators
# of classes 2 classes 3 classes 4 classes # of classes 2 classes 3 classes 4 classes
Parameters 19 26 33 Parameters 26 40 54
BIC 5083.113 5097.917 5122.462 BIC 5099.981 5142.282 5180.444
Class C1: 89.79% C1: 82.13% C1: 73.62% Class C1: 66.81% C1: 67.23% C1: 57.45%
Proportions C2: 10.21% C2: 9.36% C2: 9.36% Proportions C2: 33.19% C2: 28.51% C2: 22.55%
C3: 8.51% C3: 9.36% C3: 4.26% C3: 15.75%
C4: 7.66% C4: 4.26%
Model 2: Class-specific means and covariances between correlated indicators and class-invariant variances Model 5: Class-specific means with class-invariant variances, covariances between all indicators
# of classes 2 classes 3 classes 4 classes # of classes 2 classes 3 classes 4 classes
Parameters 21 30 39 Parameters 27 34 41
BIC 5081.717 5094.202 5118.927 BIC 5110.856 5126.486 5149.624
Class C1: 70.21% C1: 68.51% C1: 61.62% Class C1: 64.26% C1: 81.70% C1: 57.02%
Proportions C2: 29.79% C2: 24.68% C2: 28.07% Proportions C2: 35.75% C2: 9.36% C2: 25.11%
C3: 6.81% C3: 6.48% C3: 8.94% C3: 10.64%
C4: 3.83% C4: 7.23%
Model 3: Class-specific means and variances, class-invariant covariances between correlated indicators Model 6: Class-specific means, variances, class-invariant covariances between all indicators
# of classes 2 classes 3 classes 4 classes 1 class 2 classes 3 classes 4 classes
Parameters 24 36 48 20 32 44 56
BIC 5093.073 5117.181 5163.308 BIC 5124.351 5144.160 Model did not converge to an interpretable solution
Class C1: 69.97% C1: 65.16% C1: 49.97% Class C1: 74.89% C1: 70.21%
Proportions C2: 30.04% C2: 28.83% C2: 28.42% Proportions C2: 25.11% C2: 26.38%
C3: 6.02% C3: 11.22% C3: 3.40%
C4: 10.38%

Note 1: Each model additionally estimated k-1 parameters for class proportion and k-1 parameters for covariate effects

Note 2: Model selected as the final model is indicated in bold