An open-source estimation of latent class extended mixed models in R

Linear Mixed-Effects Models (LMMs) are often employed in longitudinal studies to describe changes over time of quantitative variables. Just as often, LMMs are used when the experimental dependent variable is a quantitative Gaussian, excluding cases in which the statistical population is not homogeneous. Using lcmm-package (Proust-Lima, Philipps, Diakite, & Liquet, 2016) of R–the well-known open-source language and environment for statistical computing and graphics–is possible to widen LMMs to miscellaneous/not homogeneous populations and to apply joint analysis of longitudinal and survival data on quantitative Gaussian/non Gaussian variables, therefore also considering ordinal or count outcomes. Latent class extended mixed models allow to understand whether there are unobserved groupings in a dataset. They combine the mixed models approach considering the correlation between repeated measures (random-effects) and the latent class models, identifying latent groups when searching trajectories. Moreover, the lcmm-package gives the possibility to make a set of useful post-fit functions, such as goodness-of-fit analyses, predicted trajectories of new data, posterior probabilities of group membership, latent profile-plots, individual dynamic prediction of the event, and evaluation of the predictive accuracy assessment. We used latent class extended mixed models to datasets with longitudinal measures. Our aim was to identify how and whether university students emotionally reacted to the two terroristic attacks occurred in Paris, 13th November, 2015, and in Bruxelles, 22nd March, 2016. The participants were 218 university students already involved in a micro-longitudinal study proposing online questionnaires. They completed the Emotion Regulation Questionnaire (Gross & John, 2003) focused on emotion reappraisal and suppression. They then completed daily a short version of the Achievement Emotions Adjective List (Raccanello, Brondino, & Pasini, 2014) on ten emotions. We analysed the data to identify groups of participants with similar emotional patterns during several consecutive days following the attacks, also taking into account their emotion regulation strategies.