Four data sets characterizing gross crustal waveguide configuration and attenuation properties in Eurasia (surface topography, Moho depth, sediment thickness, and Lg coda Q value) are used to examine path influences on short-period regional P/Lg ratios. Linear regressions show considerable correlations between log P/Lg and waveguide properties for both earthquake and explosion data. This is of interest because P/Lg ratios are considered to be promising regional discriminants for which it is desirable to reduce scatter caused by propagation effects. To develop a comprehensive and stable model describing the path effects, multivariate regression is applied to the data set. The waveguide properties considered are highly correlated with each other, causing collinearity in the regression process. After backward elimination, we obtain a final model with statistically significant regression coefficients, which involves two independent variables: the path attenuation term &ggr;Δ and maximum sediment thickness on each path. These two factors are associated with (1) the overall intrinsic and scattering attenuation in the waveguide and (2) the localized blockage effects caused by waveguide geometry, respectively. Using the final model to correct the observed P/Lg measurements reduces the variances for separate earthquake data and explosion data by 40% and 27%, respectively. This correction slightly enhances the performance of the discriminant for periods near 1 s. ¿ American Geophysical Union 1994