Forecasting Oil Price by Hierarchical Shrinkage in Dynamic Parameter Models
Table 8
Measures of forecast performance for log return of Brent with multivariate models.
Constant variance
Stochastic variance
MAFE
MSFE
MLPL
MAFE
MSFE
MLPL
Model (h=1)
LASSO on constant and TVPs
0.061
0.007
3.350
0.077
0.011
2.696
LASSO only on constant coeff.
0.059
0.006
3.556
0.071
0.008
3.203
LASSO only on TVPs
0.069
0.009
3.022
0.081
0.011
2.383
TVP regression model
0.072
0.008
3.037
0.077
0.010
2.953
Constant coeff. model
0.072
0.008
3.029
0.076
0.010
2.874
Model (h=12)
LASSO on constant and TVPs
0.320
0.168
0.663
0.386
0.236
0.470
LASSO only on constant coeff.
0.328
0.165
0.538
0.303
0.153
0.546
LASSO only on TVPs
0.363
0.213
0.566
0.395
0.244
0.452
TVP regression model
0.376
0.221
0.480
0.314
0.175
0.482
Constant coeff. model
0.376
0.223
0.479
0.343
0.197
0.439
Note. The value noted in bold and underlined text indicates a model performing the best out of all models, while the bold and italic text represents a model performing the worst. MSFE, MAFE, and MLPL refer to the mean squared forecast error, mean absolute forecast error, and mean log predictive likelihood, respectively.
