### Tables 8 to 9 show, for small numbers of inspectors (two to three) and each calibrated estimator, the average calibration statistics for all documents. The median RE across all combinations of all documents is shown in the second column. The third column shows the average values of k2 across documents, where k2 was estimated based on the median RE of each document. The average variance of the uncalibrated and calibrated estimates are then shown in the fourth and fifth columns. Finally, the last column shows the average increase in variance through calibration, which can be seen as an empirical estimate of k2. We can see that the change in variance predicted by equation 9 is confirmed by the data. For example, for two inspectors and Mh(JE) calibration increases the variance by a factor of 3.7021, which is close to the predicted factor of 3.6853. The increase in variance is larger for estimators with a lower cross-documents median RE. For example, for two inspectors, Mh(JE) has a lower median RE (-0.5172) than M0(MLE) (-0.3575) and therefore the increase in variance is larger (k2 decreases from 3.7021 to 2.248). In addition, we can see that the increase in variance is worse for smaller numbers of inspectors, since the average bias of the original estimate is less in those cases. The question is now whether this kind of calibration can be used in practice to obtain improved estimates, although it results in larger estimation variances.

2001

### Tables 8 to 9 show, for small numbers of inspectors (two to three) and each calibrated estimator, the average calibration statistics for all documents. The median RE across all combinations of all documents is shown in the second column. The third column shows the average values of k2 across documents, where k2 was estimated based on the median RE of each document. The average variance of the uncalibrated and calibrated estimates are then shown in the fourth and fifth columns. Finally, the last column shows the average increase in variance through calibration, which can be seen as an empirical estimate of k2. We can see that the change in variance predicted by equation 9 is confirmed by the data. For example, for two inspectors and Mh(JE) calibration increases the variance by a factor of 3.7021, which is close to the predicted factor of 3.6853. The increase in variance is larger for estimators with a lower cross-documents median RE. For example, for two inspectors, Mh(JE) has a lower median RE (-0.5172) than M0(MLE) (-0.3575) and therefore the increase in variance is larger (k2 decreases from 3.7021 to 2.248). In addition, we can see that the increase in variance is worse for smaller numbers of inspectors, since the average bias of the original estimate is less in those cases. The question is now whether this kind of calibration can be used in practice to obtain improved estimates, although it results in larger estimation variances.

2001

### Table 2 Objects identified through refinements

2001

"... In PAGE 9: ... 7 is easier to analyse and comprehend. In addition to the objects identified during the refinements (a summary is given in Table2 ), the lattice G. Canfora et al.... ..."

### Table 1. Results obtained by refining the Query in the example.

"... In PAGE 4: ... From their analysis new related queries have been extracted. For instance, from the analysis of the first repository shown above, the Query Suggester formulates the following four queries: Q1: road test approach Q2: test road approach Q3: comput test road Q4: vision test road Table1 shows the results of the search obtained through the four suggested queries. The relative values of Precision and Recall are represented in Figure 3, showing their average trends, along with the increasing number of text-units in the repository.... ..."

### Table 2 Inference rules for process extraction.

2005

Cited by 2

### Table 2: Characteristics of the corpus of anno- tated sentence pairs.

"... In PAGE 2: ... pairs of sentences were manually annotated as to the relationships between them. Table2 com- pares the characteristics of the data clusters. For each cluster, two hired judges worked in- dependently in labeling each sentence pair for Cross-document Structure Theory (CST) rela- tionships.... ..."

### Table 1. Progression of grid sizes through refinement and coarsening

1996

Cited by 17

### Table 1: Progression of grid sizes through refinement and coarsening.

1996

Cited by 10

### Table 3: MI Improvements through Mutual Refinement Strategy (MRS)

"... In PAGE 8: ... For this evaluation we measured MIs before and after MRS process. Table3... ..."

Cited by 1