Seismic hazard analysis of spatially distributed structures requires spatial correlation of the intraevent residuals of ground‐motion intensity measures (IMs). Modeling of spatially correlated intraevent residuals of IMs is influenced strongly by the structure of the correlation; specifically, whether or not the isotropic assumption is valid. The present study applied a nonparametric test using spatial statistics to evaluate the isotropic assumption of IMs. The main parameters in the anisotropic domain are the anisotropy ratio and anisotropy angle. The anisotropy ratio is determined by dividing the largest range of correlation by the smallest. The anisotropy angle gives us the direction that has the largest range of correlation. These two properties are evaluated to assess the anisotropy of the residuals of the IMs. The isotropic assumption of peak ground acceleration (PGA) and the spectral acceleration (SA) obtained from the past nine earthquakes were examined. The parameters considered were the effects of focal mechanism, regional site conditions (), earthquake magnitude, and hypocentral distance. The results show that the focal mechanism of a strike‐slip earthquake correlated well with the anisotropy angle, which can be determined by the strike direction of the focal mechanism. Testing showed that the anisotropy ratio of the PGA residuals increased as the anisotropy ratio of the and earthquake magnitude increased. The effects of and earthquake magnitude on the anisotropy ratio of the SAs were not significant. A new procedure was developed based on the results for seismic hazard analysis that considers the effect of anisotropy.