Forecasting Oil Price by Hierarchical Shrinkage in Dynamic Parameter Models
Table 7
Measures of forecast performance for log return of Brent with autoregression (AR) models.
Constant variance
Stochastic variance
MAFE
MSFE
MLPL
MAFE
MSFE
MLPL
Model (h=1)
LASSO on constant and TVPs
0.067
0.008
2.825
0.089
0.014
1.969
LASSO only on constant coeff.
0.061
0.007
3.209
0.079
0.010
2.722
LASSO only on TVPs
0.098
0.015
1.989
0.104
0.016
1.822
TVP regression model
0.090
0.013
2.299
0.100
0.015
1.960
Constant coeff. model
0.091
0.013
2.275
0.097
0.014
1.996
Model (h=12)
LASSO on constant and TVPs
0.293
0.163
0.673
0.480
0.385
0.359
LASSO only on constant coeff.
0.337
0.207
0.502
0.462
0.321
0.313
LASSO only on TVPs
0.549
0.485
0.358
0.637
0.684
0.299
TVP regression model
0.557
0.523
0.337
0.586
0.533
0.281
Constant coeff. model
0.557
0.520
0.337
0.576
0.506
0.275
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.
