In this study, we use two sets of residuals of the response spectral values to identify the repeatable effects due to specific ray paths and sites that are generally included in the standard deviation of the ground‐motion prediction equations (GMPEs): one empirical set of residuals from the Abrahamson et al. (2014)GMPE and one set derived from simulated waveforms of the CyberShake platform. Comparisons of the residuals at an oscillator period of 3 s (lower usable limit of the 3D CyberShake simulations) show similarities between the site effects estimated from the two datasets with a correlation coefficient of 0.63, but the path effects are not similar (correlation coefficient<0) for more than half of the sites evaluated. Comparing the CyberShake path effects for moderate (Mw 6.5) and large (Mw 8) earthquakes showed that, in most cases, path effects from large‐magnitude earthquakes can be approximated using path effects from smaller earthquakes that have the same closest point to a site. Attention should be used when very‐large‐magnitude (Mw 8+) events are involved and the fault crosses the sites. Finally, the method for performing a fully nonergodic site‐specific probabilistic seismic‐hazard analysis, including source, path, and site terms is outlined.