### Table 2: The aggregated counts and the true relative risks of the simulated data sets. The quantiles are given as the empirical quantiles of the simulated values.

"... In PAGE 24: ...iven in the data set used in Section 6, ranging from 3.0 to 393.1 and with a median of 19. A summary of the two simulated data sets used in the study is given in Table2 , and the realisations of B4CTDCD4B4DCCYB5B5CYBPBDBNBMBMBMBND2 and the corresponding regional relative risks, given by the mean B4C8CYBEBTCX CTDCD4B4DCCYB5B5BPD2CX over the D2CX lattice nodes within region CX, are shown in Figure 8. The prior distribution for the precision AS and the range parameter D6 are assumed to be inde- pendent.... In PAGE 26: ...ated using the block-sampling approach described in Section 3.2. As pointed out in that section, the optimal choice of block-size can be considered to be a trade-off between compu- tational cost and the acceptance probabilities of the Metropolis-Hastings steps. To study the effect of changing the block-size on the acceptance probabilities, we ran 11000 iterations of the sampler on data set I of Table2 for four different choices of blocks, keeping the hyper- parameters fixed at their true values. The blocks are made up from single regions, 1.... ..."

### Table 3: The expected (BXCX) and observed B4DDCXB5 aggregated counts, true relative risks (CACX) and SMR for region 16 and its neighbours for data set I.

### Table 4: The expected (BXCX) and observed B4DDCXB5 aggregated counts SMR for the regions for which the acceptance rates of the log-risk updates for the oral cavity cancer data are less than 10%.

"... In PAGE 32: ... The acceptance rates are reasonably high for all but a few regions, as illustrated in the bottom panels of Figure 12. The data for the regions for which the mean acceptance probabilities of the log-risk updates are less than 10% are listed in Table4 , and we observe that they all have a relatively large or small SMR or a high observed count, one of which is... ..."

### Table 6 : Consumption Risk Sharing Regressions 1982-1987 PSID Data

1999

"... In PAGE 29: ...tocks. For these two cases, one cannot reject that a is greater than zero at below the .1 level of statistical signi cance. The results presented in Table6 further dramatize the di erent propensities to share risk based on the number of alternative assets held by each household. The estimates of equation (28) presented in this table are for sub-samples of the data based on the number of types of assets owned by the household.... In PAGE 29: ...21 Ceteris Paribus one would conjecture that as a household holds an increased number of di erent assets that this should be an avenue for them to share risk with other households. The results in Table6 in fact demonstrate that, as the number of 20McCarthy (1995) nds that aggregate risk sharing improves when households are less likely to be liquidity constrained. However, he does not consider within region and industry risk sharing.... ..."

### Table 10 Summary Statistics Aggregate Indices

"... In PAGE 33: ... Risk reduction seems to be more worthwhile by combining indices from different rating and different maturity buckets: the last row in Table 9 indicates that correlations between such index-pairs are somewhat lower and less often significantly different from zero than for pairs within the same rating class or maturity bucket. A second indication of the effect of risk reduction characteristics across the rating and maturity dimensions can be found in Table10 . This table reports similar summary statistics as in Table 8 (Summary Statistics of Yield Spread Changes) but now for value weighted aggregate indices.... In PAGE 34: ...way from zero. Ljung-Box statistics significant at the 5% level are also denoted by an asterisk. Again, using the Jarque-Bera test (not shown), the normality hypothesis for the aggregate series can be rejected in every case. This can clearly be seen from the skewness and kurtosis coefficients in Table10 . Many series show significant negative skewness and, more importantly, all series show excess kurtosis.... In PAGE 34: ... The second dimension seems to be credit rating, as borne out in the lower rating categories. However, the evidence shown in Table 9 and Table10 is very crude as the aggregate rating and maturity indices are not independent from each other. It was shown in Table 4 that maturity distributions were not identical across rating classes.... In PAGE 37: ... Because these indices have very low market value (on average less than 1% of total market value, see Table 4), the fit would be even better if value-weighted averages had been used. Obviously, the correlations with the aggregate series from Table10 (not shown) are much higher due to the diversification of the unique components. We conclude that our methodology leads to a reasonable fit.... ..."

### Table 3 Aggregate Value of Banks apos; Shares and Aggregate BIF Insurance Subsidy Value loan unregulated PCA(1) PCA(2) PCA(3)

"... In PAGE 25: ... The remaining 9 banks each face a unique loan opportunity set that consists of the 9 pairwise loan combinations taken from opportunity set A, B, or C. Individual bank loan opportunity sets are enumerated in column 1 of Table3 where A1 indicates loan 1 from opportunity set A and so on. Except for di erent loan opportunity sets, all banks have the same exogenous parameter values used in the prior examples.... In PAGE 25: ... Except for di erent loan opportunity sets, all banks have the same exogenous parameter values used in the prior examples. The row entries in the second and third column of Table3 report, respectively, individual banks apos; optimal share values and the corresponding values of their deposit insurance guarantee gross of the initial premium. Similar entries in the remaining columns report bank apos;s optimal share and insurance values under a PCA with the indicated penalty rate.... In PAGE 25: ... Similar entries in the remaining columns report bank apos;s optimal share and insurance values under a PCA with the indicated penalty rate. The nal row in Table3 records the total subsidy granted this hypothetical banking system owing to under-priced deposit insurance. In the absence of any capital regulations, the hypothetical banking system would... In PAGE 26: ... The penalty rate increase from 1 to 2 reduces aggregate equity values by less than 3 (less than one-half percent) and an additional dollar increase diminishes equity prices by only about 1 in aggregate. Overall, the results in Table3 show that PCA generates substantial reductions in BIF risk and the aggregate insurance subsidy. For most banks in this example, the... ..."

Cited by 1

### Table 1: Aid, Policy and the Risk of Conflict: a Simulation

"... In PAGE 11: ... The baseline is a hypothetical country with characteristics set at the mean of all the aid-recipients in the CH sample. These characteristics are shown in the first column of Table1 . The second column of Table 1 reports the coefficients on these ... In PAGE 16: ... In turn, growth has both a direct effect on risk reduction and indirect effects via the level of income and the structure of the economy. The fourth column of Table1 collates these effects of a one point policy improvement and aggregates them into an effect on the risk of conflict. It can be directly compared with the third column which presents the baseline risk, built up from its component parts.... In PAGE 16: ... After five years, the higher income is about half as important as the direct effect of faster growth. Since the effect of primary commodity dependence upon the risk of conflict is substantial, the effect of policy improvement on risk via this route is quite large, again being shown in the fourth column of Table1 (rows 3 and 4). Because both primary commodity dependence and its square enter the logit regression, the net effect of reduced dependence is the net effect of the change in these two variables.... In PAGE 19: ...17 The overall effect of the increase in aid on the risk of conflict is shown in the fifth column of Table1 which again collates these individual effects. The risk of conflict is reduced from the baseline case of 11.... In PAGE 19: ...educed from the baseline case of 11.7% to 11.5%. In the last column of Table1 we simulate the effect of policy improvement and increased aid in combination. Because aid and policy are complements, the increased aid now has a greater effect on risk reduction.... ..."

### Table 3: Claims associated with each risk area.

1996

"... In PAGE 9: ...) Variable Entropy Gini Error F02 00 0 F01 0 0 111 F14 465 840 60 F21 668 711 52 F19 1756 1367 119 F22 2223 2177 85 F18 2258 2034 374 F12 2667 2589 289 F16 3013 2461 82 F07 3051 2947 1407 F15 3543 3122 189 Variable Entropy Gini Error F13 4629 4928 224 F20 4843 5175 19 F17 6164 6197 1023 F04 6773 6743 4041 F10 6799 6431 3975 F06 8760 9122 10195 F08 10771 11772 20842 F09 11362 10200 7895 F03 12073 12349 11464 F11 12329 12989 1480 Table 2: Attribute frequencies. Table3 records the number of claims associated with the rules generated from each of the three trees, aggregated by the number of claims. There are 1090 rules from the Entropy tree, for exam- ple, which have just a single claim.... ..."

Cited by 6

### Table 3: Claims associated with each risk area.

1996

"... In PAGE 9: ...) Variable Entropy Gini Error F02 0 0 0 F01 0 0 111 F14 465 840 60 F21 668 711 52 F19 1756 1367 119 F22 2223 2177 85 F18 2258 2034 374 F12 2667 2589 289 F16 3013 2461 82 F07 3051 2947 1407 F15 3543 3122 189 Variable Entropy Gini Error F13 4629 4928 224 F20 4843 5175 19 F17 6164 6197 1023 F04 6773 6743 4041 F10 6799 6431 3975 F06 8760 9122 10195 F08 10771 11772 20842 F09 11362 10200 7895 F03 12073 12349 11464 F11 12329 12989 1480 Table 2: Attribute frequencies. Table3 records the number of claims associated with the rules generated from each of the three trees, aggregated by the number of claims. There are 1090 rules from the Entropy tree, for exam- ple, which have just a single claim.... ..."

Cited by 6

### Table 6 Predictability and Predictiveness of Spending classified by Aggregated Condition Categories 1996 and 1997 Medicare data

"... In PAGE 17: ... Among those that are not predictable, several have relatively high and plausible predictiveness: Septicemia/Shock , Diabetes with No complications , and Other Infectious Diseases . Results by Aggregated Condition Categories (ACCs) Results using diagnoses and summing up spending by Aggregated Condition Categories (ACCs) from the DxCG risk adjustment system are presented in Table6 and Figure 4. Rather than sorting them by predictive power, we have left them sorted by ACC to see illustrate how the classification system is organized.... ..."