### Table 6: The name disambiguation accuracy, mean and standard deviation on 9 DBLP datasets of different names, using multiple schemes of attributes with both the naive Bayes approach(Bayes) and the SVM approach(SVM); and the statistical significance (two tail P value) of the performance difference by the two approaches.

"... In PAGE 7: ... The columns Training/Test size list the number of citations in the training/test dataset. Table6 shows the performance on the 9 DBLP name sets by both approaches. Because of the different citation data quality and prob- ability distributions these datasets have from the J Anderson and J Smith datasets, both approaches achieve different performance than on the previous two datasets.... ..."

### Table 3. Tests of Differences Between Meansa: Absolute values of t-statistics and p-values for Two-Tail Tests.

in Experimenting On Classroom Experiments: Do They Increase Learning In Introductory Microeconomics?

"... In PAGE 9: ... Within the experimental group, students in the section receiving a grade incentive for profits improved their TUCE scores by slightly more than students given no grade incentive. Table3 presents results of testing for differences between sub-sample means of variables used in the study. Results in the middle column of Table 3 pertain to differences between the control section and the two experimental sections combined, while results in the right-hand column pertain to differences between the two experimental sections, according to whether or not a grade incentive was offered for profits earned in experimental markets.... In PAGE 9: ... Table 3 presents results of testing for differences between sub-sample means of variables used in the study. Results in the middle column of Table3 pertain to differences between the control section and the two experimental sections combined, while results in the right-hand column pertain to differences between the two experimental sections, according to whether or not a grade incentive was offered for profits earned in experimental markets. The t-statistics presented in Table 3 can be computed from information in Table 2 assuming unknown but equal... In PAGE 9: ... Results in the middle column of Table 3 pertain to differences between the control section and the two experimental sections combined, while results in the right-hand column pertain to differences between the two experimental sections, according to whether or not a grade incentive was offered for profits earned in experimental markets. The t-statistics presented in Table3 can be computed from information in Table 2 assuming unknown but equal... In PAGE 10: ...As shown in Table3 , the difference between mean changes in TUCE scores in experimental and control groups is statistically significant at the five percent level in a two-tail test. The null hypothesis that variances of the change in test score are equal between experimental and control sections is not rejected at the five percent level, although it would be rejected at the ten percent level.... In PAGE 20: ... As mentioned previously, however, exams in each section covered course material as presented in that section. Although the information is not presented in Table3 , there were no significant differences in final exam scores between any of the sections. Yet another test might be based on student evaluations of instructor effectiveness, but these were not conducted in Spring 1999 as part of a transition to a new administration.... ..."

### Table 1: Impact of the heavy-tailed session length and think time distributions

2003

"... In PAGE 8: ... The mean session inter-arrival time for the workloads is chosen such that UWebCPU is 77%. Table1 summarizes the results. Impact of heavy-tailed session length distributions To isolate the effect of heavy-tailed session length distributions, we compare system performance with the Empirical workload and the Heavy-tailed session length workload.... In PAGE 8: ... From Figure 2 (a), the cumulative distribution of response time for the Heavy-tailed session length workload exhibits a longer tail since it covers a wider spread of response time values than the Empirical workload. From Table1 , the Heavy-tailed session length workload yields a higher Rmean and R95 than the Empirical workload even though the system was operated at the same UWebCPU (77%) for both workloads. The R95 value with the Heavy-tailed session length workload is 1.... In PAGE 9: ...Figure 2: Effect of the heavy-tailed session length distribution Impact of heavy-tailed think time distributions Next, we compare the Empirical workload and the Heavy-tailed think time workload to study the impact of heavy-tailed think time distributions. From Table1 , both Rmean and R95 for this workload are greater than those of the Empirical workload confirming the findings reported in [3] that longer think times degrade the responsiveness of the system. The reason for this behaviour can be explained by observing that heavy-tailed think times within sessions also cause long-lasting sessions similar to heavy-tailed session lengths.... In PAGE 9: ... Noting that the mean session arrival rate is the same for both workloads, this leads to increased request bursts at the system that can degrade response time, as explained in the previous paragraph. However, from Table1 , comparing the R95 ratio of 1.17 for the Heavy-tailed think time workload with the corresponding ratio of 1.... In PAGE 9: ...ession length workload is 1.47 times that for the Empirical workload. However, the degree of sensitivity depends on the extent of additional burstiness the increase in tail density is able to induce. For example, from Table1 , we can infer that the increase in burstiness due to the increase in tail density from the empirical think time distribution to the bounded Pareto think ... ..."

Cited by 1

### Table 2. The correlation analysis results between individual structural turbulence items and the overall level of the competition (N=42)

"... In PAGE 4: ... The mean value of these three items was considered to represent the overall level of competition. In Appendix Table2 , we have presented the results of the correlation analysis between the overall and individual competition items.... ..."

### Table 3: Failures recorded at each noise level. algorithm, however, some of these \failures quot; do eventually converge to a solution1, but for the majority of runs, and to keep simulation times to reasonable levels, the limit was set. The number of failures at each noise level are shown in Table 3. It is clear from Fig. 11 that the scatter is not random - it is super cially a Poisson distribution, with a long exponential tail. Taking a crude mean value is therefore meaningful qualitatively, if not perhaps mathematically optimal. Fig. 12 shows the mean training time as a function of injected synaptic noise level. A small but statistically signi cant reduction in training

in Enhanced MLP Performance and Fault Tolerance Resulting from Synaptic Weight Noise During Training

1994

Cited by 15

### Table 3: Packet length distributions for considered arrival processes Percentage of Arrival Arrival Arrival Arrival

"... In PAGE 21: ... In varying these four distributions, the mass of the packet length distribution was shifted from favoring short packets to favoring long packets. The four packet length distributions are shown in Table3 . The mean utilization of the arrival process to each input queue remained xed at 10% for all experiments.... ..."

### Table 1 displays the performance extrema for this deterministic proof by contra- diction strategy on the testbed as well as the mean values over all successful runs. The values in brackets indicate the deviation from the mean. Fig. 2 shows the underlying dis- tribution of the run time for these experiments. In fact, the distribution exhibits heavy- tailed behavior [2] which is manifested in the long tail of the distribution stretching for several orders of magnitude.

2001

"... In PAGE 4: ... Table1 . Statistics for successful runs (108 out of 160) on testbed using deterministic strategy.... ..."

Cited by 1

### Table 1: Mean Hop Count ( h) and Mean Message Delay ( d) for di erent values of . (n gt; 85700 messages)

1994

"... In PAGE 20: ... The price paid for the ability to circumvent a highly utilized network area is an increase in mean path length h. The means of path length and message delay for di erent values of are summarized in the Table1 . Figures 5 and 6 show the corresponding graphs for the d and h respectively.... ..."

Cited by 7

### Table 1: Mean Hop Count ( h) and Mean Message Delay ( d) for di erent values of . (n gt; 85700 messages)

1994

"... In PAGE 20: ... The price paid for the ability to circumvent a highly utilized network area is an increase in mean path length h. The means of path length and message delay for di erent values of are summarized in the Table1 . Figures 5 and 6 show the corresponding graphs for the d and h respectively.... ..."

Cited by 7

### Table 1: Correlations of Delay Activity in Subintervals of ISI

"... In PAGE 19: ... This value is compared with the MRCs of extreme ISI intervals in trial subsets of similar statistics, with balanced subsets of low and high averages. Those would correspond to the first and second row in Table1 . In fact, the numbers are close.... ..."