Assessing the application of latent class and latent profile analysis for evaluating the construct validity of complex posttraumatic stress disorder: cautions and limitations

Figures & data

Table 1. Latent class analysis models and fit indices.

Model	Log-likelihood	# estimated parameters	df	Entropy	BIC	SA-BIC	AIC	BLRT p-value*
2 classes	−996.60	23	222	.86	2119.72	2046.81	2039.19	<.001
3 classes	−969.56	35	210	.91	2131.67	2020.72	2009.12	<.001
4 classes	−954.30	47	198	.90	2167.17	2018.18	2002.61	.012
5 classes	−939.38	59	186	.92	2203.34	2016.32	1996.78	.042

Df = degrees of freedom; BIC = Bayesian Information Criterion; SA-BIC = Sample-size adjusted BIC; AIC = Akaike Information Criterion; BLRT = bootstrap likelihood ratio test. * BLRT values could not be calculated in poLCA and where therefore computed following an analogous analysis in Latent Gold.

Figure 1. Symptom endorsement patterns for the 2-class model following latent class analysis. Symptom abbreviations: RE = re-experiencing; AV = avoidance; HY = hyperarousal; AD = affective dysregulation; NSC = negative self-concept; DIR: disturbances in interpersonal relationships.

Figure 2. Symptom endorsement patterns for the 3-class model following latent class analysis. Symptom abbreviations: RE = re-experiencing; AV = avoidance; HY = hyperarousal; AD = affective dysregulation; NSC = negative self-concept; DIR: disturbances in interpersonal relationships.

Table 2. Latent profile analysis models and fit indices.

Model	Log-likelihood	Entropy	BIC	ICL	BLRT p-value
2 classes	-3868.36	.58	-7923.77	-7953.77	.001
3 classes	-3802.96	.81	-7858.98	-7903.06	.001
4 classes	-3762.29	.81	-7843.64	-7903.74	.001
5 classes	-3738.21	.80	-7861.51	-7942.18	.001
6 classes	-3720.80	.81	-7892.69	-7975.74	.017
7 classes	-3709.84	.80	-7936.79	-8026.62	.36

Abbreviations: BIC = Bayesian Information Criterion; BLRT = bootstrap likelihood ratio test; ICL = integrated complete-data likelihood criterion.

Figure 3. Class proportions and mean scores on each item for each class, for the 4-class model (with spherical clusters of unequal size) resulting from the latent profile analysis. Symptom abbreviations: RE = re-experiencing; AV = avoidance; HY = hyperarousal; AD = affective dysregulation; NSC = negative self-concept; DIR: disturbances in interpersonal relationships.

Figure 4. Jittered scatterplot with the sum of present PTSD symptoms (items scoring >1 on a 0–4 scale) versus the sum of present CPTSD-specific items (items scoring >1 on a 0–4 scale). Colour of points corresponds to class membership of the three-class model following latent class analysis; (blue) diamond = class 1 (CPTSD with low symptom endorsement), (red) square = class 2 (CPTSD), (grey) triangle = class 3 (PTSD).

Figure 5. Jittered scatterplot with the sum of the scores on all six PTSD symptom items (scored 1–5) versus the sum of the scores on all CPTSD-specific items (scored 1–5) that were included in the latent profile analysis. Colour of points corresponds to class membership of the four-class model following latent profile analysis; (blue) diamond = class 1, (red) square = class 2, (grey) triangle = class 3 (PTSD), (orange) cross = class 4.