After the 2011 Fukushima accident, the seismic regulations for nuclear power plants (NPP) in Japan have been strengthened to include countermeasures far beyond design-basis accidents. The importance of seismic probabilistic risk assessments, therefore, have been the focus of deserved attention. Generally, an uncertainty quantification has been a very important undertaking to assess for fragility in NPP buildings. Therefore, this study focuses on the reduction in epistemic uncertainty by aiming to clarify the seismic-response effects on NPP buildings based on different modeling methods. As a first step in this study, the authors compared the seismic-response effects using two modeling methods of analysis. To evaluate the seismic response, an analysis was performed on two building model types; these being the three-dimensional (3D) finite-element model and the sway-rocking model with a conventional lumped mass system. To input a ground motion, the authors adopted 200 types of simulated seismic ground motions, generated by fault-rupture models, using stochastic seismic source characteristics. For the uncertainty quantification, we conducted a statistical analysis of the seismic responses acquired from the two modeling methods based on the building response each ground-motion input, and quantitatively evaluated the uncertainty response by considering the different modeling methods. We found a clear difference in the modeling methods near the floor and wall openings. We also imparted our knowledge on these 3D effects for the seismic-response analysis.