Abstract: In this study, 4695 strong-motion records from subduction slab earthquakes in Japan obtained by the K-NET and KiK-net networks were used to develop a damping modification factor （DMF） model for the vertical acceleration spectrum. The DMF model considering the damping ratio and spectral period is established for four site classes, and the geometric mean values for DMF are used to perform regression fitting in respect to the damping ratio and spectral period by least squares method. To evaluate the influence of hypocenter, path, and site effects on the residual distribution of the model, the total model residuals and standard deviations were separated into between-event and within-event parts, and the within-event residuals were further divided into the between-site and within-site parts by using a random effects model. The results show that the effect of damping ratios on DMF can be modelled by the third-order polynomial of the logarithm of the damping ratios, a fourth-order polynomial of the logarithm of spectral periods can be used to model the effect of spectral periods for the DMF. The effects of source and path parameters were not modelled in this study because of DMF model is used to scale a design spectrum not associated with a given scenario earthquake. The distribution of the between- and within-event residual distributions with respect to moment magnitude, fault depth and source distance suggests that including these terms would improve prediction capability of the model when a DMF is designed for scaling a 5% acceleration spectrum associated with a scenario event for a give magnitude and source distance.