- Ground-motion model is:
where Y is in cm∕s2; α

_{1}= -1.7918, α_{2}= 1.61, α_{3}= -1.00, α_{4}= -0.0058 and σ = 0.60 for CU; α_{1}= 0.4089, α_{2}= 1.23, α_{3}= -1.00, α_{4}= -0.0016 and σ = 0.55 for SCT; α_{1}= 0.4656, α_{2}= 1.22, α_{3}= -1.00, α_{4}= -0.0012 and σ = 0.58 for CDAO. - Use data from 3 stations in hill (Ciudad Universitaria, CU, 22 records) and lake-bed (Secretaría de Comunicaciones y Transportes, SCT, 15 records, and Central de Abastos, CDAO, 13 records) zones of Mexico City.
- Focal depths 40.0 ≤ H ≤ 128.4km.
- Data quite well distributed w.r.t. M
_{w}. - Most data from r
_{rup}< 300km. - Use Bayesian linear regression (Ordaz et al., 1994) because of limited data. Prior probability distributions of coefficients are obtained from empirical model of source spectrum with 2 corner frequencies, frequency-dependent attenuation parameters for the region, duration estimates and random vibration theory. For the lake-bed stations use this information plus the 1D analytical transfer function using soil profiles.
- Compare final coefficients to prior estimates.
- Examine residuals w.r.t. M
_{w}and r_{rup}and find no clear trends. - Try including H in model but insufficient data to constrain coefficient.
- Compare predicted and observed spectra for all data. Find good match except for a few poorly-recorded events.