Virtual Reality (VR) technology represents a new medium to provide immersive solutions in different fields. The analysis of a user while interacting in VR, through data science and machine learning (ML) techniques, might provide insights to deliver customized functionalities that enhance productivity and efficiency in learning tasks in education or rehabilitation processes in healthcare. However, empirical research involving VR often borrows methods from human-computer interaction intending to evaluate human behavior through technology, whereas ML intend to create mathematical models, usually with non-empirical approach. Their opposite nature might cause confusion for early-stage researchers wanting to understand and follow the methodological approaches and communicative practices in empirical studies that merge both VR and ML. This paper presents a scoping review of methodological strategies undertaken in 21 peer-reviewed research articles that involve both VR and ML. Results show and appraise different methodological approaches in research projects, and outline a set of recommendations to combine metrics from inferential statistics and evaluation of ML models to increase validity, reliability and trustworthiness in future research projects that intersect VR and ML.