Supplementary file 2
Description of the LC model and explanation how the expected values in the figure were calculated from the model parameter estimates
The basic idea of the latent class (LC) model is that associations between manifest items arise because the population is composed of a finite number of mutually exclusive and exhaustive unobserved (latent) classes that are characterised by different distributions of the manifest indicators. As a model-based clustering approach, LC analysis accounts for measurement error in parameter estimation. The decision on the most appropriate number of LCs is based on likelihood fit indices. Furthermore, given an individual’s response pattern with regard to the manifest indicators, it is possible to determine probabilities of membership for the different LCs.
A figure depicting the expected values of the identified LCs for the nine indicators was produced to characterise the LCs in accordance with the selected model (see Figure 1). These expected values were computed by summing the products of the category numbers (from 3 = “very important” to 0 = “not at all important”) and the corresponding conditional (LC specific) response probabilities over the four categories of the respective indicator and then transforming them to fit on a 0–1 scale. The plot was produced using the freely available statistical software package R [55], version 3.1.0.