Stochastic reconstruction (SR) is one of the most usable molecular reconstruction approaches applicable to heavy petroleum cuts, using statistical distribution to describe the structure attribute of molecules in a mixture. Suitable values for distribution parameters are estimated during an optimization process on the base of mixture characterization data. This study focuses on searching the connection between the parameter of structure attribute distributions and different types of experimental data, used for hydrocarbon mixture characterization. An algorithm has been introduced that allows to determine which types of analytical data affect the distribution of structure attributes based on testing statistical hypotheses. The core idea is looking for a significant correlation between calculated properties and distribution parameters, and then comparing the magnitude of parameter impact with measurement precision. It was shown that different types of analytical data can affect different parameters of the reconstructed composition: distillation curve data is the main source of information about distributions variance, while elemental analysis and PNA helps determine mean values for various structure attributes; information concerning the number of branches in molecules are strongly connected with C-13 NMR data.