Choosing Transformations for Quarterly Time Series Regression

Summary

The document raises a modeling question about regressing quarterly time series when the response and predictors use different growth horizons. In particular, it asks whether year-over-year changes in explanatory variables can be used to predict quarter-over-quarter changes in an outcome, or whether both sides should use the same transformation.

It provides no proposed model, empirical evidence, or cited literature, so it does not resolve the question. The choice depends on the research objective and the series’ dynamics; differing horizons can be meaningful but may affect interpretation, serial dependence, and alignment. The document is best treated as a prompt to examine stationarity, timing, and predictive validity rather than as evidence for a particular transformation.

Key ideas

• The document asks whether regression variables need to use matching growth horizons.
• It considers using annual changes in predictors to explain quarterly changes in a response.
• It offers no answer, empirical analysis, or literature references to support either choice.
• Transformation choices should be evaluated in light of timing, time-series properties, and the intended interpretation.

Tags

# Transforming Variables in time series regression


# Transforming Variables in time series regression












I have multiple quarterly time series data and trying to build a linear regression model using this dataset.

- Should the transformations on the LHS and RHS be the same i.e QoQ percent changes?

- Could we have Qi/Qi-4 percentage changes on the independent variables to predict QoQ (Qi/Qi-1) changes in response variable?

- If yes, is there any literature or book that illustrates or refutes this fact.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.