This paper describes the statistical analysis of data on coke formation (measured by microcarbon residue) in five resids, as a function of compound classes, in order to identify variables with a significant effect on the former. In this sense, the SAS VARCLUS and Principal Component Analysis were used. The analysis carried out leads to a lineal combination of the H/C ratio, vanadium and nitrogen content as a model to explain coke yield, under non-catalytic conditions, as determined by MCR. Of all these factors, the H/C ratio has the most important contribution.