Skip to main content
. Author manuscript; available in PMC: 2024 Jan 1.
Published in final edited form as: J Head Trauma Rehabil. 2022 Dec 30;38(4):E254–E266. doi: 10.1097/HTR.0000000000000848

Table 2.

Model Fit Statistics for Gray Matter and White Matter Latent Profile Analyses.

Classes Class Proportions Loglikelihood AIC BIC SABIC BLRT p BLRT LMR p LMR Entropy
GM Model (n = 156)
2 0.56 / 0.44 −8539.163 17188.33 17356.07 17181.98 619.91 p<.001 613.52 .002 .933
3 0.33 / 0.49 / 0.19 −8435.054 17018.11 17243.80 17009.56 208.22 p<.001 206.07 .337 .933
a4 0.21 / 0.38 / 0.22 / 0.19 −8340.402 16866.80 17150.44 16856.07 189.30 p<.001 187.35 .194 .947
5 Model not well-identified
WM Model (n = 144)
2 0.54 / 0.46 −8419.603 16961.21 17142.37 16949.35 548.95 p<.001 543.74 .003 .922
3 0.51 / 0.33 / 0.15 −8316.315 16796.63 17040.16 16780.69 206.58 p<.001 204.62 .178 .933
a4 0.19 / 0.19 / 0.24 / 0.38 −8231.739 16669.48 16975.37 16649.45 169.15 p<.001 167.55 .417 .939
5 0.16 / 0.23 / 0.33 / 0.15 / 0.13 −8168.981 16585.96 16954.22 16561.85 125.51 p<.001 124.32 .348 .956
6 Model not replicated

Note: GM = Gray matter; WM = White matter; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion; SABIC = sample-size adjusted BIC; BLRT = Bootstrapped Likelihood Ratio Test; LMR = Lo-Mendell-Rubin Likelihood.

Lower values of AIC, BIC, SABIC demonstrate better model fit. pBLRT < .05 indicates that the k-class model is a better fit to the data than the k-1 class model. pLMR small p-values indicate that the k-class model fits better to data than the k-1 class model. Entropy value close to 1 indicates excellent classification of subjects into latent classes.

a

Denotes the selected models. Although the 5-class WM model (under WM model) is the best fit model regarding the fit indices, it does not demonstrate the appropriate separation of the classes compared to 4-class model.