Autoregressive to anything: Time-series input processes for simulation (Q2564301)

We present a model for representing a stationary time-series input process \(\{Y_t\); \(t=1,2,\dots\}\) with an arbitrary marginal distribution and any feasible autocorrelation structure specified through lag \(p\). We use a transformation-oriented approach to represent \(\{Y_t\}\). This approach takes a process with a known autocorrelation structure, the base process \(\{Z_t\}\), and transforms it to achieve the desired marginal distribution for the input process, \(\{Y_t\}\). The target autocorrelation structure of \(\{Y_t\}\) is obtained by adjusting the autocorrelation structure of the base process. In our model, the base process is a standardized Gaussian autoregressive process of order \(p\) \((\text{AR}(p))\), so we refer to \(\{Y_t\}\) as an ARTA (autoregressive to anything) process.