Fig. 3: Forecasting error and uncertainty decline when including exogenous variables.
From: Adaptive food price forecasting improves public information in times of rapid economic change
Mean absolute error (MAE) (a) and average prediction interval width (b) by selection algorithm and the month of the forecast, 2003–2024. The difference is reported between selection algorithms based on including all exogenous variables (kitchen-sink) or none (benchmark).
