A Note on Regressions with Integrated Variables

The purpose of this note is to contribute to modeling time series that may be characterized as integrated. In particular, a discussion about the use of standard distributions in regression models is included and some essential concepts are previously reviewed to this effect. The main objective is to analyze, through examples, the transformations required to formulate the model with stationary variables, emphasizing the fact that these transformations are not necessarily required. Likewise, several cases of practical interest are presented, such as the representation in levels or in differences in the Granger causality analysis and the autoregressive distributed lag models.