Grain size distribution is one of the paleoenvironmental proxies that provide insight statistical distribution of size fractions within the sediments. Multivariate statistics have been used to investigate the depositional process from the grain size distribution. Still, the direct application of the standard multivariate methods is not straightforward and can yield misleading interpretations due to the compositional nature of the raw grain size data. This paper is a methodological framework for grain size data characterization through the centered log ratio transformation and euclidean data, coupled with principal component analysis, cluster analysis, and linear discriminant analysis to examine Quaternary sediments from Tövises bed in the southeast Great Hungarian Plain. These approaches provide statistically significant and sedimentologically interpretable results for both datasets. However, the details by which they supplemented the conceptual model were significantly different, and this discrepancy resulted in a different temporal model of the depositional history.