- Ground-motion model is:
where Y is in m∕s2, a = -1.817, b = 0.460, c = -1.428, d = 0.271 and σ = 0.417.

- Use 3 site classes in final model:
- s = -1
- Negative residuals from first regression
- s = 0
- Zero residuals from first regression
- s = 1
- Positive residuals from first regression

Originally use 4 geological units from a macro-zoning for correction of observations to rock:

- Q
- Quaternary: very soft soils and alluvium gravel deposits (V
_{s,30}≤ 360m∕s - V
- Volcanic: quaternary-tertiary age (360 ≤ V
_{s,30}≤ 1000m∕s) - T
- Tertiary: sediments, soft rocks and flysch deposit, 360 ≤ V
_{s,30}≤ 800m∕s - M
- Mesozoic: carbonate platform succession, V
_{s,30}> 800m∕s

- Use data from 21 stations of Irpinia Seismic Network, a high dynamic-range, dense, seismographic network (co-located Geotech S-13J and CMG-5T instruments or a Trillium sensor), from 09/2005 onwards.
- Develop for use in ShakeMaps for small earthquakes.
- Exclude earthquakes from outside area of interest (120 × 120km
^{2}) and events with M_{L}< 1.5. - De-trend and band-pass filter (0.075–20Hz, 4-pole Butterworth) records. Retain records with signal-to-noise ratio > 5 and where PGA occurs during 5–95% Arias-intensity interval.
- Reasonably uniform distribution w.r.t. M and R, although only one event with M
_{L}> 2.9. - Use three-step approach. Firstly adjust PGAs to rock based on geological classification. Next derive model using these adjusted PGAs. Examine residuals for each station and perform a Z-test with 5% significant level to find sites with mean residuals significantly different from zero. Finally add site term to account for these average differences.
- Test inclusion of quadratic-magnitude term but find that not significantly different from zero.
- Test inclusion of anelastic attenuation term but find that not significantly different from zero.
- Find that the inclusion of the site term d reduces σ and that this reduction is significant by using the Akaike Information Criterion.
- Compare observed and predicted (grouped by magnitude ranges) PGAs by distance and via residual plots and find good match.