A note on regressions with integrated variables

Abstract

The purpose of this note is to contribute to the modeling of time series that can be characterized as integrated. In particular, the issue of using standard distributions in regression models is discussed. To do this, some essential concepts are previously reviewed. The main objective is to analyze through examples the transformations necessary to formulate the model with stationary variables, emphasizing that these transformations do not necessarily have to be carried out. Different cases of practical interest are also presented, such as representation in levels or differences in the analysis of non-causality in the Granger sense and autoregressive models of distributed lags.

JEL classification: C22