A Class of Models for Non‐normal Time Series

Abstract.We propose a model for non‐normal times series in which we regard the series as an instantaneous transformation of an underlying ‘generating’ series which is normal. We describe a procedure which simultaneously estimates both the transformation to normality and the time series structure of the underlying series. This model has several advantages; in particular several alternative types of forecast can easily be calculated. The relative merits of these forecasts are considered.