Motivated by a burnout syndrome dataset which belongs to the RN4CAST project, we propose and analyse graded response models, based on the item response theory (IRT) to deal with polytomous response items. Our proposal is to compare estimation via marginal maximum likelihood, the Monte Carlo Markov chain via Gibbs sampler and the Monte Carlo Markov chain via NUTS methods. The best performance for the item discrimination parameter estimation was obtained by the marginal maximum likelihood method, whereas both Monte Carlo Markov chain methods produce good results for the item difficulty parameters estimation. In general, the NUTS method showed to be the most indicated in the individual latent traits estimation.


