Understanding the time-dependent behaviour of a nuclear reactor following an intended or unintended change of the reactor conditions is of crucial importance to the safe operation of nuclear reactors. Safety evaluations of nuclear reactors involve the analysis of normal and abnormal reactor operation states with adequate and validated simulation tools. There is a growing interest in supplementing the results of best-estimate reactor calculations with uncertainties. A major source of uncertainty in reactor calculations is the nuclear data. The propagation of uncertainties allows to study their impact on response values of reactor calculations and to establish safety margins in nuclear reactor safety evaluations. For the simulation of reactor transients, the two-step approach consisting of the generation of group constants in infinite lattice geometry and the full-core calculation using a neutron-kinetic/thermal-hydraulic code system has been state-of-the-art. Nowadays, Monte Carlo codes allow for the determination of group constants based on a three-dimensional full-core model. Thus, effects from neighbouring fuel assemblies can be considered. In this thesis, the applicability of the neutron-kinetic code DYN3D coupled with the thermal-hydraulic system code ATHLET for the simulation of transients in small reactor cores was assessed. Experimental reactivity accident tests carried out at the SPERT III research reactor were simulated. Appropriate group constants were generated with the Monte Carlo code Serpent. The overall good agreement between the calculated and the experimental results prove the applicability of the applied codes for the analysis of such reactor transients. Nuclear data uncertainties were propagated through the simulations of two control rod withdrawal transients in a mini-core model of a pressurised water reactor. The random sampling-based method XSUSA in combination with the SCALE code package was used to determine varied two-group constants. Based on these two-group constants, 1,000 DYN3D-ATHLET calculations were performed and statistically analysed to obtain uncertainty information for the reactor power and the reactivity. Since the assumption of normality could not be maintained for the reactor power, distribution-free Wilks tolerance limits were applied. By sensitivity analyses, the number of delayed neutrons per fission of U-235 and the scattering cross sections of U-238 were identified as the major contributors to the reactor power uncertainty. For a potential improvement of transient simulations, an approach was developed that was comprised of the determination of group constants using a Serpent full-core model at selected times during a transient simulation and the updating of the original infinite lattice group constants with these full-core group constants. Various challenges and possible solutions were identified. The applicability of the group constants obtained by a full-core model was assessed by their application in full-core diffusion calculations.