In this comment, I consider the issue of serial dependence in the Probability Integral Transforms that arises when performing multi-step ahead forecasts. I provide a simple technique for correcting the autocorrelation that relies on fitting a Moving Average under correct specification. I show via simulation and empirically that this technique appears reliable. Lien vers l'article
CHEVILLON, G. (2014). Multi-step Forecast Error Corrections: A Comment on “Evaluating Predictive Densities of US Output Growth and Inflation in a Large Macroeconomic Data Set” by Barbara Rossi and Tatevik Sekhposyan. International Journal of Forecasting, 30(3), pp. 683-687.