Hello Claramae,

a latent class model as binary indicators; a latent profile model has quantitative indicators. I would nonetheless refer to the latent variables as latent classes as they are categorical.

HTH

Holger

Typically, we refer to the analysis as (classical) latent class analysis (LCA) when categorical (i.e., nominal, binary, or ordinal) indicators (items) are used. In classical LCA, the item parameters are the conditional (class-specific) response probabilities for the different item response categories. When the latent class indicator variables (items) are continuous (metrical, interval scale), the item parameters are the class-specific means and variances. With continuous indicators, the analysis is often referred to as latent profile analysis (LPA).

In some software programs such as Mplus, you can have a mixture of categorical and continuous indicators in the same model so that the distinction between classical LCA and LPA becomes blurred. In your case (5-point Likert scale), I would probably go with LCA (treating the indicators as categorical/ordinal) rather than treating the indicators as continuous. You do not need to dichotomize your variables. You can treat them as polytomous-ordinal in LCA and estimate conditional response probabilities for all categories.

The reason I would prefer LCA is because it is properly designed to handle ordinal variables, whereas LPA is for continuous (metrical) variables. 5-point Likert items are not truly continuous. In addition, LCA is a "full-information" approach that uses the entire information from the observed item response pattern frequencies in parameter estimation, whereas in LPA "only" summary statistics (means, variances, covariances) are used.

Latent class analysis is categorical latent variables measured with categorical items, while latent profile analysis is measured with continuous items. You either use the formula in Excel to analyze your dataset or use R, SAS, MPLUS, etc. All the best.