Table 8

Indicator-level out-of-sample prediction

ModelOutcomePLS RMSELM RMSEΔRMSEPLS MAELM MAELower PLS RMSE
M1FAIR0.7600.772−0.0120.5700.5782/2
M1ACC0.7680.784−0.0160.5780.5893/3
M1TRANS0.7300.766−0.0360.5420.5673/3
M1TRUST0.6900.692−0.0020.5180.5131/3
M1Overall0.7350.752−0.0170.5510.5609/11
M2AFAIR0.7610.769−0.0080.5730.5792/2
M2AACC0.7720.784−0.0120.5810.5893/3
M2ATRANS0.7330.759−0.0260.5480.5643/3
M2ATRUST0.6910.697−0.0060.5190.5162/3
M2AOverall0.7370.751−0.0130.5540.56110/11
M2BFAIR0.7640.769−0.0050.5740.5771/2
M2BACC0.7740.794−0.0200.5820.5983/3
M2BTRANS0.7420.771−0.0280.5480.5703/3
M2BTRUST0.6930.701−0.0080.5200.5182/3
M2BOverall0.7420.758−0.0160.5540.5659/11

Note(s): Ten-fold PLSpredict with ten repetitions. Values are averages across the indicators belonging to each outcome. LM = linear-model benchmark; ΔRMSE = PLS RMSE minus LM RMSE, so negative values favour PLS. The final column reports the number of indicators for which PLS produced a lower RMSE than the benchmark

or Create an Account

Close subscription notice
Close access options