Forecasting Oil Price by Hierarchical Shrinkage in Dynamic Parameter Models
Table 9
Measures of forecast performance for log return of Brent (h = 1) with UCSV models.
Constant variance
Stochastic variance
MAFE
MSFE
MLPL
MAFE
MSFE
MLPL
Model (h=1)
LASSO on constant and TVPs
0.064
0.007
3.111
0.076
0.010
2.969
LASSO only on constant coeff.
0.060
0.007
3.229
0.066
0.007
3.474
LASSO only on TVPs
0.069
0.009
2.879
0.075
0.010
2.750
TVP regression model
0.070
0.008
2.898
0.069
0.008
3.071
Constant coeff. model
0.070
0.008
2.909
0.068
0.008
3.148
Model (h=12)
LASSO on constant and TVPs
0.280
0.137
0.807
0.283
0.138
0.629
LASSO only on constant coeff.
0.297
0.157
0.530
0.269
0.117
0.774
LASSO only on TVPs
0.281
0.137
0.778
0.275
0.135
0.768
TVP regression model
0.298
0.160
0.511
0.276
0.121
0.706
Constant coeff. model
0.301
0.160
0.517
0.283
0.124
0.799
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.
