Selection of an appropriate ground‐motion prediction equation (GMPE) is a key element within seismic‐hazard analysis (SHA). A new methodology is introduced in this paper in order to assess the stability of GMPEs. The proposed methodology, resampling analysis (RSA), evaluates the sensitivity of GMPEs under a given subset of resampled data. The model bias is calculated in the proposed methodology on the basis of the statistical hypothesis tests for different residual components. Four Next Generation Attenuation models were evaluated in order to investigate their stability by means of statistical RSA within their own databases. The case study results show that some of the considered GMPEs are quite sensitive to their own databases. Hence, the RSA methodology, as a stability criterion, is proposed as a practical tool within the GMPE development and also as an effective and complementary tool for selection of the most appropriate GMPE within an SHA.
