Abstract
Ground-motion prediction equations (GMPEs) are developed by regression analysis based on datasets of recorded ground-motion parameters at multiple stations during different earthquakes and in various source regions. The variability of the GMPE is considerable and its influence on probabilistic seismic hazard analysis is paramount.Much recent work has examined the distance-dependent variance of PGA and PGV and estimated the standard deviation increases with decreasing distance. In contrast, some studies have estimated the standard deviation decreases with decreasing distance. A possible reason of this distance-dependence is the heavy weight for short distances in regression analysis. In this paper, the influence of weighted regression analysis and a diverse set of strong motion data (data consisting of only strike-slip earthquakes and mixed data of reverse and strike-slip earthquakes) on distance-dependent variance of PGA and PGV is examined. When we do not apply weights during the regression analysis, the standard deviations does not clearly show the distance dependent variance. Added to this, the distance-dependent variance can't be clearly seen when we use the mixed data set.