### Table 2: Characteristics of evolved monolithic GP solutions for 1000 out-of-sample random evaluations in D = [?2:0;2:0].

"... In PAGE 7: ... It evolves a monolithic expression that achieves the desired behavior. Table2 compares two best GP(IF) solutions obtained with two GP solu- tions in the simple representation. The errors achieved in the monolithic IF-based approach are clearly superior -2.... ..."

### Table 6: Out-of-sample forecasts

1995

"... In PAGE 20: ... This evidence pro long memory is also supported by out-of-sample forecasts. Table6 shows the relative improvement or deterioration of the MSE of the ARFIMA(0; d; 0) speci cation to the MSE of the random walk with drift for a 6- and a 12-month forecast horizon for weekly, monthly and quarterly exchange rate changes. Except for the highly unreliable results of quarterly data, all ARFIMA predictions show a slight improvement over the random walk with drift model.... ..."

Cited by 3

### Table 19: Out-of-sample R

1994

"... In PAGE 11: ... Interestingly, results from the subperiods in uenced by the October 1987 crash still yield lower prediction er- rors for the learning networks than for the Black-Scholes model, except for near-term in-the-money options. For completeness we also show the out-of-sample R 2 apos;s [see Table19 ] and the absolute hedging error compari- son [see Table 20] as we did in Section 3.4 for the syn- thetic data.... In PAGE 11: ...4 for the syn- thetic data. Table19 , for instance, shows that the aver- age out-of-sample R 2 of roughly 85% for the estimated Black-Scholes model is somewhat worse than that of the other network models. Note however that unlikethe case for our synthetic data, the options in the S amp;P 500 data set are not independent, and thus wemust look at these results with caution.... ..."

Cited by 67

### Table 19: Out-of-sample R

1994

"... In PAGE 11: ... Interestingly, results from the subperiods in uenced by the October 1987 crash still yield lower prediction er- rors for the learning networks than for the Black-Scholes model, except for near-term in-the-money options. For completeness we also show the out-of-sample R 2 apos;s [see Table19 ] and the absolute hedging error compari- son [see Table 20] as we did in Section 3.4 for the syn- thetic data.... In PAGE 11: ...4 for the syn- thetic data. Table19 , for instance, shows that the aver- age out-of-sample R 2 of roughly 85% for the estimated Black-Scholes model is somewhat worse than that of the other network models. Note however that unlikethe case for our synthetic data, the options in the S amp;P 500 data set are not independent, and thus wemust look at these results with caution.... ..."

Cited by 67

### Table 17: Estimation results, out-of-sample forecasts

2002

"... In PAGE 30: ... The reason for restricting the model to the six key variables is to obtain a comparison between the in-sample and the out-of-sample performance of our benchmark model. Table17 shows the estimation results for the various sub-samples that are used to generate the out-of-sample forecasts. The 1997 Asian Crisis: Out-of-sample forecast in December 1996 For predicting the Asian crisis, Table 17 shows that both the short-term debt to reserves ratio and the current account deficit lose their significance if the model is estimated ending in December 1996.... In PAGE 30: ... Table 17 shows the estimation results for the various sub-samples that are used to generate the out-of-sample forecasts. The 1997 Asian Crisis: Out-of-sample forecast in December 1996 For predicting the Asian crisis, Table17 shows that both the short-term debt to reserves ratio and the current account deficit lose their significance if the model is estimated ending in December 1996. However, these results are intuitively convincing: it is well know that a high level of short term debt was one of the key factors that had made Asian countries vulnerable in 1997.... ..."

Cited by 2

### Table 5: Out-of-sample variances of Returns

"... In PAGE 7: ... Hence, the forecasted hedge ratio is the one-period forecast of the conditional covariance divided by the one- period forecast of the conditional variance. These out-of-sample estimated variances of returns are reported in Table5 , while the out-of-sample percentage variance reductions are reported in Table 6. Our results show that the hedge ratio9 obtained from OLS and ECM generates better results in terms of hedging effectiveness.... ..."

### Table 4. Out-of-Sample Prediction Performance

"... In PAGE 20: ... The combinations estimators generally achieve better performance than the LAD and OLS estimators but the magnitude of improvement is very small. Prediction performance measured by prediction 2 R is summarized in Table4 . The prediction 2 R is not necessarily positive because out-of-sample predictions are not guaranteed to be orthogonal to out- of-sample residuals.... In PAGE 20: ... Therefore, a positive prediction 2 R indicates that the predictor is better in terms of PMSE than the sample mean assumed known in advance. According to the summary statistics in Table4 , the return on ADC TeleCom Co. is more difficult to predict than that for HomeStake Co.... ..."

### Table 1: Out-of-sample forecast performance

"... In PAGE 13: ... However, for all other variables actual values are used. The root mean square errors (RMSE) for out-of-sample forecasts (1996:4 { 1997:3) for one to four quarters ahead are presented in Table1 . With the exception of the rst quarter, predictions of the monetary model in its structural form compare most favorably to those based on more restricted equations.... ..."

### Table 3 Out-of-sample simulation results

"... In PAGE 28: ...Figure 5 depicts (in the middle) the out-of-sample simulations of wealth over time and (at the bottom) the resulting marginal distribution of wealth at the end of the in- vestment horizon. Table3 summarizes the statistics of wealth obtained at the end of the investment horizon. The mean wealth is $1.... In PAGE 36: ...1.006 million. Thus, the target wealth of $1 million is exceeded with larger than 95% probability. The statistics in Table3 give a certainty equivalent wealth of $998,000 as the lowest of all four utility functions, reflecting again the low emphasis on the upside displayed by the quadratic downside utility function. 7.... ..."

### Table 1 Out-of-Sample Average Hedge Ratios

2002

"... In PAGE 11: ... 9 experiment repeated until 73 forecasts for each model are generated. Table1 provides a summary of the hedge ratios. Likewise, evaluations of hedge ratio performance are reported in Table 2.... In PAGE 11: ... Also, because new data are added each week the OLS hedge ratio, like the SUR and MGARCH methodologies, will experience some variability (albeit small) in the out-of-sample analysis. Results presented in Table1 illustrate the average hedge ratios generated for all three econometric specifications (OLS, SUR and MGARCH) for each of three trader models. It is clear that the average hedge ratios for the underlying commodity generated from the MGARCH and SUR methodologies differ from those generated by OLS.... ..."

Cited by 1