BAYESIAN ATTENUATION REGRESSIONS - AN APPLICATION TO MEXICO-CITY

We describe the application of a bayesian linear regression technique to the problem of deriving strong-motion attenuation relations. This a pproach provides a conceptual framework for the formal incorporation o f knowledge about the involved phenomena that comes from sources other than the observed data (prior information, according to the bayesian terminology). The procedure produces numerical solutions that are more stable and rational than those obtained from conventional regression schemes. We illustrate the use of the proposed technique with the deri vation of attenuation laws for the Fourier acceleration spectrum, as a function of magnitude and distance, at a hill-zone station in Mexico City.