### Table 2: Comparing the three transmission channels bank lending channel {5.3

1996

"... In PAGE 29: ...5. Results for the tests of H1, H2,andH3are shown in Table2 . Since the null hypotheses of no e ect have one-sided alternatives, critical values for a one-tailed test will be used to evaluate the three hypotheses.... In PAGE 29: ... Again the test of the bank lending channel has a non-standard distribution because the parameter u is unidenti ed under H1. According to the evidence in Table2 , the hypothesis of no bank lending channel, H1,canbe rejected. The hypothesis of no broad credit channel, H2, and the hypothesis of no asymmetric channels, H3, cannot be rejected.... ..."

Cited by 2

### Table 4 presents, for each of the above models and for each of the 48 states j, the posterior median estimate of j and the (one-sided) p-value of the actual election result actual j

1998

"... In PAGE 17: ... If the p-values for a model are clustered near 0 or 1, the forecasts are overcon dent (that is, the standard errors are too small); if they are clustered near the middle of the range, the forecasts are undercon dent and the standard errors too large. This behavior is summarized in the bottom rows of Table4 by the scaled mean z-score and sum of squares of the normal-transformed p-values. The sum of squares can itself be compared to a 2 distribution, and this reveals that the models with (a2; b2) set to (1; 1) (that is, Rj 0:5... In PAGE 24: ...999 0.997 Table4 : Posterior medians and p-values (Pr( j actualjdata,model)) for each state for each of 8 models. At the bottom of the table are a measure of t (mean absolute error of state estimates) and three measures of calibration (scaled mean Z-score of state p-values, sum of squares of the Z-scores, and p-value of the sum of squares compared to the 2 48 distribution).... ..."

Cited by 2

### Table 2 : One-sided t-test results

2005

"... In PAGE 5: ...able 1 : Platform order.................................................................................................... 12 Table2 : One-sided t-test results .... ..."

### Table 2. Comparison of timings for baseline function evaluation and derivative computation using one-sided nite di erences, forward mode, and the iterated reverse mode with a single processor, CFL3D 5.0. Machine Timings (seconds) Ratio to Function

1998

"... In PAGE 6: ...ives computed for CFL3D 5.0 and CFL3D 4.1 have not yet been investigated. Table2 presents timings based on CFL3D 5.0, including the shape analysis procedure, nite dif- ferences, ADIFOR-generated forward-mode derivative {Executing CFL3D 5.... In PAGE 6: ... No- tice that the cost of computing all 88 derivatives us- ing ADJIFOR on the SUN SMP is about 10 percent of the cost of using nite di erences. The timing numbers shown in Table2 is intended to illustrate the performance of ADJIFOR-generated code rela- tive to other standard methods for computing deriva- tives. As such, they do not re ect the possible use of coarse-grained parallelization within nite di erences or forward-mode derivative computations to improve performance.... ..."

Cited by 18

### Table 4: Problem 6.1, one-sided preconditioner

in Eigenvalue

"... In PAGE 27: ... (13) 465 11 12 1719 13 14 6603 16 18 25875 20 22 102435 29 32 Table 5: Problem 6.2 The behaviour of the one-sided preconditioner in (13) is illustrated in Table4 and Table 5, third column. Built with a second order factorization D1 of M (with drop tolerance 0:001) and an incomplete factorization of the Schur complement (with drop tolerance 0:01) first approximated with the matrix C + K, this preconditioner shows also very good iteration counts, although somewhat higher than for the two-sided method.... ..."

### Table 2: Average errors of various methods for all one-sided

2000

Cited by 65

### Table 4: Average errors of various methods for all one-sided

2000

Cited by 65

### Table 7: One-sided matching to nearest neighbor

1999

"... In PAGE 5: ... A resulting data set with local recoding is shown in Table 7. In Table7 each row corresponds to at least two households in the population. Therefore the one-sided nearest neighbor local recoding can withstand the fishing strategy attack ... In PAGE 6: ... For example household No.9 of the hypothetical population corresponds only to the 9th row of Table7 and in this sense the 9th row of Table 7 might be identified. This weakness clearly results from the one-sidedness of the matching.... In PAGE 6: ... For example household No.9 of the hypothetical population corresponds only to the 9th row of Table 7 and in this sense the 9th row of Table7 might be identified. This weakness clearly results from the one-sidedness of the matching.... In PAGE 7: ... Age Size Income Occup. 1 30-40 4 400 A 2 50 2-3 700-800 B 3 30-40 4 400 A 4 40 4-5 600 C 5 50 2-3 700-800 B 6 30 3 300-500 A 7 40 4-5 600 C 8 50 2-6 500 A 9 50 2-6 500 A 10 30 3 300-500 A From Table7 and Table 8 we see that two-sided nearest neighbor local recoding leads to stronger protection accompanied with larger average distance within pairs. The advantage of the two-sided nearest neighbor local recoding is that it withstands both the fishing... ..."

Cited by 1