
260 Current Trends in Bayesian Methodology with Applications
OLS
AIC
BIC
LASSO
HPM
MPM
BMA
SPIKE
50
100
150
200
FIGURE 12.4
Boxplot of test errors for the Prestige dataset.
12.4.5 Analysis of the Cars dataset
The cars dataset c omes from the R package datasets. The cars dataset has
n = 80 and p = 4. Here we try to predict the gas milea ge (MPG) of a car
given four characteristics, namely, VOL, HP, SP and WT. We spot from Figure
12.5 that no method emerges a s predictively superior for this dataset.
12.4.6 Analysis of the Hald Cement dataset
The Ha ld Cement dataset comes from the R package MPV. This dataset with
n = 13 and p = 4 is clearly the smallest dataset