Time Deformation, Continuous Euler Processes and Forecasting

. A continuous Euler model has time‐varying coefficients. Through a logarithmic time transformation, a continuous Euler model can be transformed to a continuous autoregressive (AR) model. By using the continuous Kalman filtering through the Laplace method, this article explores the data application of a continuous Euler process. This time deformation of an Euler process deforms specific time‐variant (non‐stationary) behaviour to time‐invariant (stationary) data on the deformed time scale. With these time‐invariant data on the transformed time scale, one may use traditional tools to conduct parameter estimation and forecasts. The obtained results can then be transformed back to the original time scale. Simulated data and actual data such as bat echolocation and the US residential investment growth are used to demonstrate the usefulness of time deformation in forecasting. The results indicate that fitting a traditional autoregressive moving‐average (ARMA) model on an Euler data set without imposing time transformation leads to forecasts that are out of phase while the forecasts of an Euler model stay mostly in phase.