Exploratory Data Analysis July 19, 2004 NGA Workshop
Step 1: For each earthquake, fit data to the following model (functional form used by Sadigh et al.) M = Earthquake magnitude Rj = Closest distance to fault All soil types are lumped together
C5 and C6 can’t be reliably estimated for most earthquakes fixed to the values of Sadigh et al. (1997) 172 earthquakes 172 x 2 = 344 unknown coefficients Rj < 70 km Freefield records California data vs. worldwide data
Step 2: Examine estimated C1 and C4 for magnitude effects and functional form.
Step 3: Modify the basic model to C1 C1 + C2 M + C3 (M-8.5)2.5 C4 C4 + C5 M Repeat regression
Step 4: Examine estimated coefficients for Magnitude scaling at short distances to large-magnitude earthquakes Style of faulting effect Effects of coseismic surface faulting
Step 4: Examine residuals for Site classification and site effects FW/HW effects Directivity effects
Observed trend may be explained by more than one explanatory variables Z1.5 ~ Vs30 (NEHRP) Style of faulting ~ Surface Rupture FW/HW ~ Directivity
Y Cos()
No Yes Unknown SS 17 17 38 NM 3 5 14 RV 18 4 11 RV-OB 7 3 1 NM-OB 1 1 4 Unknown 0 0 28
Exploratory Outliers do show up