### Table 5: Price Model Estimates with Robust t-Statistics

### Table 4 presents estimates of four linear probability models, with t statistics computed using robust 9

"... In PAGE 12: ... We find statistical confirmation of the self-reported fact that most economists do 13 not weigh costs and benefits by testing the joint hypothesis that the first five regressors have zero 14 coefficients. The second-to-last row of Table4 shows p-values for that hypothesis, which reveal the 15 weak predictive power of subjective costs and benefits in the first two models. This weak predictive 16 power does not result from overall weakness of the prediction equation, however, as likelihood ratio 17 tests of the hypothesis that all coefficients in the model are zero are easily rejected across all models.... In PAGE 13: ...13 mismatch). 1 2 Statistical predictors of the PSA recommendation? 3 The dependent variable in the fourth model presented in Table4 is a forced-choice survey item asking 4 for recommendations about PSA screening for an asymptomatic man in his 50s. Recall that 5 unconditional rates of recommendation were almost twice the rate of PSA taking, 85 versus 46 percent.... In PAGE 26: ...14 0.03 Table4 : Estimated linear probability models for the PSA decision and recommendation *H0 is the joint hypothesis that the first five variables, which proxy for perceived costs and benefits, have zero effect on the probability of having (or recommending) a PSA. The test statistic is distributed as F(5, sample size minus number of regressors) under the null.... ..."

### Table 6 Robustness Tests of the State Fixed Effect Models Note: T-statistics appear in parentheses. The regressions are identical to models #3 and #4 in Tables 3 and 4 except as noted in the rows indicating added regressors and changes in sample.

in Educational Crowding Out: Do Immigrants Affect the Educational Attainment of American Minorities?

"... In PAGE 22: ...13 I then repeated models #3 and #4 from Tables 3 and 4 for blacks and Hispanics, adding the mean pupil-teacher ratio to the models which already contain controls for education and income among the older generation of the given race, as well as a fixed effect for each state. The results appear in the top panel of Table6 . Several conclusions emerge from the table.... In PAGE 25: ... This has the effect of representing both years equally in the regression sample. The results appear in the bottom panel of Table6 (regressions #5-8). About one third of the observations are lost due to the sample restriction.... ..."

### Table 6 Robustness Tests of the State Fixed Effect Models Note: T-statistics appear in parentheses. The regressions are identical to models #3 and #4 in Tables 3 and 4 except as noted in the rows indicating added regressors and changes in sample.

in Educational Crowding Out: Do Immigrants Affect the Educational Attainment of American Minorities?

"... In PAGE 22: ...13 I then repeated models #3 and #4 from Tables 3 and 4 for blacks and Hispanics, adding the mean pupil-teacher ratio to the models which already contain controls for education and income among the older generation of the given race, as well as a fixed effect for each state. The results appear in the top panel of Table6 . Several conclusions emerge from the table.... In PAGE 25: ... This has the effect of representing both years equally in the regression sample. The results appear in the bottom panel of Table6 (regressions #5-8). About one third of the observations are lost due to the sample restriction.... ..."

### Table 4: Two robustness checks Reaction function Allow an additional

1996

"... In PAGE 34: ...the \reference model quot; from Table 1 with the modi ed reaction function. The results of the robustness check are shown in the rst column of Table4 . While the di eren- tial e ect of monetary policy contractions on real GDP when z u declines with the change in the reaction function, from an additional 7.... ..."

Cited by 2

### Table 6: GARCH ML Estimation Results Model 1 Model 2 Model 3 Model 4

1998

"... In PAGE 19: ... We scale the news count variables by a factor of 0:01 as is indicated in table 6. Table6 presents Quasi-Maximum Likelihood Estimates for four different versions of our augmented GARCH(1,1) model. Standard errors from the estimated Hessian are reported in parentheses and robust standard errors (Bollerslev and Wooldridge 1992) are shown in square brackets.... In PAGE 19: ... Finally, in addition to the interest rate differential, we include currency-specific news-count variables in the variance (Model 4). Table 7 summarizes some analyses of the standardized residuals from the models re- ported in Table6 . In particular, we report descriptive statistics for the first four moments of the standardized residuals.... ..."

Cited by 5

### Table 1. As we can see from this table, finite state method is better in the relative processing speed and for processing existing patterns while statistical model is better for processing new sentence patterns, which makes the system very robust.

"... In PAGE 5: ...Fixed Fast 100% Partial Result Single FSM Fixed Fast 100% Die Stat Model Open Slow 85%(pre) 95%(rec) 75%(pre) 92%(rec) Table1 . Results Comparison On the other hand, we investigate the performance of statistical model in more detail on the blind test.... ..."

### Table 1. Comparison of CONTRAlign variants. We counted the number of times each variant out- performed or was outperformed by the basic model, and assigned p-values using a simple yet robust statistical sign test to check for deviations from a symmetric distribution in which either aligner is equally likely to do better. Accuracy improvements relative to the basic model are signiflcant in every case with the exceptions of the local and PSIPRED single sequence prediction models.

2006

"... In PAGE 9: ...2 Comparison of model topologies and feature sets In our flrst set of cross-validation experiments, we selected each of the reference databases in turn as the testing set, and used alignments pooled from the other three databases as the training set.3 Table1 compares the various models de- scribed in Section 2.3 as evaluated on each of the four databases.... In PAGE 11: ...s shown, cross-validated CONTRAlign (i.e., CONTRAlignHYDROPATHY) beats current state-of-the-art methods by 3-4% despite (1) estimating all model parameters, including the emission matrix, and (2) following a rigorous cross- validated training procedure. Based on the comparison of the hydropathy and basic models in Table1 , it is clear that these accuracy gains result directly from the use of hydropathy-based gap scoring. Perhaps most striking, however, is that a variety of existing methods, including CLUSTALW and MUSCLE, already in- corporate hydropathy-based modiflcations in their alignment scoring, yet do not manage to achieve above 70% accuracy on our benchmarks.... ..."

Cited by 7

### Table 1. Comparison of CONTRAlign variants. We counted the number of times each variant outperformed or was outperformed by the basic model, and assigned p-values using a simple yet robust statistical sign test to check for deviations from a symmetric distribution in which either aligner is equally likely to do better. Accuracy improve- ments relative to the basic model are significant in every case with the exceptions of the local and PSIPRED single sequence prediction models.

2006

"... In PAGE 9: ...2 Comparison of Model Topologies and Feature Sets In our first set of cross-validation experiments, we selected each of the reference databases in turn as the testing set, and used alignments pooled from the other three databases as the training set.3 Table1 compares the various models de- scribed in Section 2.3 as evaluated on each of the four databases.... In PAGE 11: ...38 n/a parameters, including the emission matrix, and (2) following a rigorous cross- validated training procedure. Based on the comparison of the hydropathy and basic models in Table1 , it is clear that these accuracy gains result directly from the use of hydropathy-based gap scoring. Perhaps most striking, however, is that a variety of existing methods, including CLUSTALW and MUSCLE, already in- corporate hydropathy-based modifications in their alignment scoring, yet do not manage to achieve above 70% accuracy on our benchmarks.... ..."

Cited by 7

### Table 4: Gravity Model Including Initial Conditions

2004

"... In PAGE 17: ... Table4 reports the results when 1995 trade is included in the regression equation for other years. As expected, the statistical significance of population, distance and some other control variables are not longer robust across all time periods.... ..."

Cited by 1