This repository contains the datasets and code used in the study “Predicting translation counterparts in causative constructions.” The datasets consist of annotated examples of Italian and Spanish causative constructions and their Czech translation counterparts. The repository includes (i) full annotated datasets for Italian and Spanish, (ii) revised datasets used for statistical modelling, and (iii) the R script used to estimate Bayesian multinomial regression models using the brms package (Stan backend). The models estimate the probability of selecting a Czech translation counterpart (TYPE) as a function of verb valency (VALENCY) and complement class (COMP_CLASS), with random effects for VERB and TRANSLATOR. The repository also contains summaries of the fitted models.