Abstract: Burnup calculation is an important component of reactor physics calculations, which provides the evolution of the isotopic inventory during the reactor operation. Reliable prediction of nuclide density with information of uncertainty is of importance for reactor operation and safety, waste transport and management. Nuclear data are the basic input parameters for the reactor physics calculation. With the improvement of calculation models and computer technology, the nuclear-data uncertainties become the main uncertainty sources for the burnup calculation. In order to perform the uncertainty propagations from nuclear cross sections to the responses of burnup calculations, two categories of methodologies have been widely applied: the deterministic method and the statistical sampling method. In this paper, a code was developed to calculate burnup calculation uncertainty caused by nuclear data based on statistical sampling method. There are three main steps to perform uncertainty propagation from the uncertainties of input parameters to the responses: firstly, determine the distribution ranges of input parameters; secondly, generate the samples of the input parameters; finally, statistically calculate for responses of corresponding input samples. Uncertainties coming from nuclear data including cross sections, decay constants and fission yields were propagated to the nuclide densities of SF95-4 sample in the Takahama-3 reactor with this newly developed code. 300 stand-alone burnup calculations were performed, each time with a different perturbed library. The 56-group cross-section covariance information from SCALE6.2 was propagated and the results show that this contribution most affect the uncertainties of actinides. Then the decay constant uncertainties from ENDF/B-Ⅷ.0 was propagated. The decay data uncertainties have inappreciable impacts on the uncertainty of nuclide densities. This effect is negligible for all actinides and most of the fission products with the exception of 151Eu, of which the 9.0% of uncertainty reflects the uncertainty of 8.9% on the decay constant of its direct parent 151Sm. Finally, the uncertainties of the fission yields were propagated and we can conclude that the uncertainties on the fission yields are the main source of uncertainty of fission products, ranging up to almost 25%. The fission yield data were taken from the original ENDF/B-Ⅷ.0 library and they were given without correlations. We generated fission yield correlations for ²³⁵U and ²³⁹Pu thermal systems using a simplified generalized least-squares method and propagated the new covariance matrices in the same model. The comparison between the inventory calculation results obtained by sampling the correlated and non-correlated ENDF/B-Ⅷ.0 fission yields shows that the introduction of correlations significantly reduce the effect of the fission yield uncertainties.