The spatial variability of soil elastic modulus not only affects the characteristics of input seismic waves but also increases the randomness of structure dynamic response and affects the reliability of structure in terms of considering the SSI effect. In order to explore the influence of the spatially variable soil elastic modulus on the seismic response of isolated structures， the random field simulation technology is combined with finite element analysis to conduct the dynamic "non-intrusive" random finite element calculation based on Monte Carlo Simulation. Taking the shaking table test of a typical large chassis isolation structure as an example， the deterministic and random finite element analyses are carried out. The relevant response parameters of the structure are calculated，and the AIC value is used to determine the best-fit marginal distribution of the response parameters. The results reveal that when the spatial variability of soil is considered， the amplification effect of non-homogeneous soft soil on the amplitude of the low-frequency seismic wave is more obvious than that of homogeneous soft soil. The peak value of the Fourier spectrum increases by about 14.2% and the distribution of spectral value concentrates to low-frequency. The deterministic analysis underestimates the amplification impacts of SSI effects on the acceleration of the isolation layer and the displacement response of floors above the isolation layer. According to the AIC value， it is not always reliable to use normal or lognormal distribution to describe the distribution of data. It needs to determine the best-fit marginal distribution of response parameters， so as to provide a more reasonable design basis for structural isolation.