- Ground-motion model is:
_{h}= 6.5; h = 5km and c_{2}= 0 (fixed due to instabilities in regression due to lack of near-source data); c_{3}= 0 (fixed because either statistically insignificant or positive). Do not report coefficients only show graphs of predictions. - Data from K-Net and KiK-Net from May 1996 to October 2011. 38 226 are borehole records from KiK-Net (581 stations) and 79 876 are surface records (1411 stations).
- Select data from events recorded by ≥ 3 stations and sites that have recorded ≥ 3 events.
- Focal depths from about 0.4 to about 30km with most between 5 and 20km.
- Data well distributed w.r.t. M
_{w}and for R > 10km. - Separate variability into components describing between sequence, between event, between site-class and between station using a mixed-effects regression technique.
- Do not aim to develop a model to calculate hazard but to understand between-event variability in terms of stress (drop) parameter. Hence aim for a simple model that does not include effects that affect only a small subset of data (e.g. nonlinear site response and directivity).
- Examine residuals and find no substantial bias. Some slight overprediction at R ≥ 180km, underprediction for 5–15km and overprediction for very short distances.