It is a special issue to model an earthquake ground‐motion relation (GMR; ground‐motion prediction equation) for the seismic‐hazard analysis of regions with low seismicity. A region‐specific GMR may be difficult to develop if there are only a few observations. Therefore, different approaches, criteria, and procedures have been developed in seismology and earthquake engineering to apply appropriate GMRs from one region to another. This is a special type of statistical model development. We criticize this and corresponding research from the statistical point of view. Our evaluation criteria result from state‐of‐the‐art mathematical statistical analysis, especially regression analysis.