Đang chuẩn bị liên kết để tải về tài liệu:
Handbook of Economic Forecasting part 54

Handbook of Economic Forecasting part 54. Research on forecasting methods has made important progress over recent years and these developments are brought together in the Handbook of Economic Forecasting. The handbook covers developments in how forecasts are constructed based on multivariate time-series models, dynamic factor models, nonlinear models and combination methods. The handbook also includes chapters on forecast evaluation, including evaluation of point forecasts and probability forecasts and contains chapters on survey forecasts and volatility forecasts. Areas of applications of forecasts covered in the handbook include economics, finance and marketing | 504 H. White Table 8 Artificial data Modified nonlinear least squares - Logistic Summary goodness of fit Hidden units Estimation MSE CV MSE Hold-out MSE Estimation R-squared CV R-squared Hold-out R-squared 0 1.30098 1.58077 0.99298 0.23196 0.06679 0.06664 1 1.30013 1.49201 0.99851 0.23247 0.11919 0.06144 2 1.30000 1.50625 1.00046 0.23255 0.11079 0.05961 3 0.91397 1.10375 0.84768 0.46044 0.34840 0.20321 4 0.86988 1.05591 0.80838 0.48647 0.37665 0.24016 5 0.85581 1.03175 0.80328 0.49478 0.39091 0.24495 6 0.85010 1.01461 0.80021 0.49815 0.40102 0.24783 7 0.84517 1.00845 0.79558 0.50105 0.40466 0.25219 8 0.83541 1.00419 0.75910 0.50681 0.40718 0.28648 9 0.80738 1.07768 0.75882 0.52336 0.36379 0.28674 10 0.79669 1.03882 0.73159 0.52967 0.38673 0.31233 11 0.79664 1.04495 0.73181 0.52971 0.38312 0.31213 12 0.79629 1.05454 0.72912 0.52991 0.37745 0.31466 13 0.79465 1.06053 0.72675 0.53088 0.37392 0.31688 14 0.78551 1.04599 0.71959 0.53628 0.38250 0.32361 15 0.78360 1.07676 0.72182 0.53740 0.36433 0.32152 16 0.76828 1.09929 0.70041 0.54645 0.35103 0.34165 17 0.76311 1.08872 0.70466 0.54950 0.35727 0.33765 18 0.76169 1.11237 0.70764 0.55034 0.34332 0.33484 19 0.76160 1.13083 0.70768 0.55039 0.33242 0.33481 20 . 0.76135 . 1.13034 . 0.70736 . 0.55054 . 0.33271 . 0.33511 . . . 41 . 0.68366 . 1.14326 . 0.65124 . 0.59640 . 0.32508 . 0.38786a results as good as those seen in Table 9. Nevertheless we observe quite good performance. The best CV MSE performance occurs with 50 hidden units corresponding to a respectable hold-out R2 of 0.471. Moreover CV MSE appears to be trending downward suggesting that additional terms could further improve performance. Table 11 shows analogous results for the polynomial version of QuickNet. Again we see that additional polynomial terms do not improve in-sample fit as rapidly as do the ANN terms. We also again see the extremely erratic behavior of CV MSE arising from precisely the same source as before rendering CV MSE useless for polynomial model .