Temporal Disaggregation, Missing Observations, Outliers, and Forecasting: A Unifying Non-Model-Based Procedure

Abstract

We suggest a simple non-model-based procedure to recover a time series from its temporally aggregated realizations. If additional assumptions on the underlying process are introduced, it is shown that the procedure is related to many of the former proposals in the literature. It can also be easily modified to deal with the estimation of missing observations and outliers, and with forecasting. Finally, some important identification issues are discussed.