"... In PAGE 21: ... However, our limited sample of forecast data prevents us from estimating a multi-dimensional nonlinear model with any accuracy. Table7 reports results from regressing the pricing errors on information about forecasts through... ..."

"... In PAGE 17: ... The mean interpurchase time ranges from 5 to 10 weeks. Note that marketing instruments in Table3 are averaged over weeks, stores and difierent UPCs. The display and feature variables therefore take values between 0 and 1.... In PAGE 21: ... To assess the impact of such a promotion, we have to rely on simulation. We use the estimated model to simulate purchases for a number of households starting at calendar time t0, where we have a promotion from t1 to t2, and for t lt; t1 and t gt; t2, we set the marketing instruments to their sample mean, see Table3 . The size of the promotion is set to one standard deviation.... ..."

"... In PAGE 5: ... 5. FEATURE EXTRACTION Table2 summarizes the features extracted for each PNR, sorted by the information gain computed for each feature. Information gain [8, 11] is a popular metric for measuring the contribution of a feature to determination of a class label.... In PAGE 5: ...5 passenger models, while ProbE accepts the continuous inputs directly, with discretization handled internally. Table2 shows that whether a passenger is ticketed, and membership in a frequent flier program have the highest in- formation gain. Although not explicitly shown in this table, ticketed passengers, as well as frequent fliers, are signifi- cantly more likely to show for a flight.... In PAGE 5: ... We expected that passengers making fre- quent itinerary changes would be more likely to not show, but this premise is not supported by the information gain. All 19 features shown in Table2 were used as input to the passenger-level models, since, as noted above, informa- tion gain neglects possible interaction with other features. The same features, with the exception of the flight destina- tion, were provided as initial input to the cabin-level models, along with the output probabilities from the C4.... In PAGE 6: ... In other words, the historical model can be viewed as an extremely bushy decision tree generated by hand. Separate APMR passenger-level models were built in each of 5 distinct segments formed from splits over the important features Passenger Ticketed, Frequent Flier, and PNR Originator shown in Table2 . These segments were de- termined via limited experimention to produce the minimum loss over the training set.... ..."

"... In PAGE 21: ..., 1996) and to quantify decomposition rates of the different soil organic matter fractions (Hassink, 1995). Fluxes Carbon fluxes at the micro-scale can be measured using static chambers installed in a field or using pot experiments ( Table6 ). Within these closed chambers, CO2 is accumulating during a preset time and the concentration is measured using a CO2 analyzer.... In PAGE 21: ... Provided that much information on the effects of the selected management measures on C sequestration to feed existing or future models is missing, measurements have to be made to collect these data. The methods men- tioned in Table6 can be used to measure C stocks in soils that result from both natural variation, indirect or direct effects. ... ..."

"... In PAGE 8: ...lain or Garden, and Number of Moves: 7 or 9, as discussed in Section 3.2. (c) Explanatory Variables These were recorded from the questionnaires and during the experiment. The major ones are recorded in Table2 . The most important ones are given earlier in the table.... In PAGE 10: ... The null hypothesis is equivalent to the subjects simply guessing moves at random, rather than based on their gained understanding of the spatial layout and the moves themselves. The independent variables (immersion, environment) and each of the explanatory variables of Table2 were considered in the analysis. The results are shown in Table 4, and the null hypothesis is rejected.... ..."

"... In PAGE 22: ...22 Table2 summarizes our data for each rule. [Insert table 2 about here] Explanatory variables always refer to the first prior to the year for which the binary indicator was calculated.... ..."

"... In PAGE 17: ...RNKPLN represents TRUNKS per line. RSLDBPLN represents RESOLDB per line. RSLDRPLN represents RESOLDR per line. Table2 describes the data for the explanatory variables. I use the price per month for a 2-wire local line in urban areas, UNEPRC, as the representative UNE price.... ..."

"... In PAGE 13: ... Of course it is preferable to include variable that have economic interpretations. In Table4 we present and explain the computation of the explanatory variables associated with the parameters #0C and #11 of equations #283#29 and #286#29. It is important to note that variables are lagged with respect to the trade time.... In PAGE 25: ...03 0.60 Table4 gives the estimates of the trade equation de#0Cned as above. Several di- agnostics are given.... ..."

"... In PAGE 5: ... In sum, except for income and price (bid variable), which are automatically included in all the regres- sions, every variable available from the USDA survey that was significant in at least one regression was included in the regressions, subject to the proviso that the variable make some sense from a farm management standpoint. Table2 presents sample statistics for these variables for all the farmers in the sample. 6 The survey procedures in place did not allow a more complex allocation of bids.... ..."