"... In PAGE 17: ...3 Validation of Analytic Analysis and Simulation The analytic model has been validated against the simulation model. Table1 compares the analytic and simulation results, where c;2 = 50 m and u;1 = u;2 = 1000 m. The table indicates that for exponential VLR residence time with m = 0:2, the error rate is less than 0.... ..."

"... In PAGE 46: ... Where memory bandwidth is the bottleneck, computa- tion rate increases accordingly. Table3 compares performance of ordered accesses as calculated analytically and mea- sured via simulation. Recall that analytic results represent a lower bound.... In PAGE 48: ... Table3 compares effective memory bandwidth for ordered accesses as calculated analyti- cally and measured via simulation. Once again, analytic results represent a lower bound on performance.... ..."

"... In PAGE 4: ... To start, consider the simplest confusion matrix where the only discrimination is between target and non-target. The confusion matrix for this simple case is shown in Table3 . In a simulation, the numbers in the confusion matrix are used to determine the outcome of a random draw each time an object, target or otherwise, is encountered.... In PAGE 8: ... The results of the simula- tion were compared to the analytical formulations discussed earlier. Table3 shows the comparison results. There is good agreement between analytical prediction and the simulation results.... ..."

"... In PAGE 12: ...xponentially distributed with rates 5.0 and 10.0, respectively, model two processors of the system. Because, all timed activities of this system are exponentially distributed, this model is Markovian and could be solved analytically. The results of its analytic solution and simulation are shown in Table2 . Simulation has been run for 10000 and 100000 units of time for before and after steady state, respectively.... ..."

"... In PAGE 33: ... Note that for LL-24 only a single vector is referenced so that no reordering is performed. Table3 compares performance of ordered accesses for the computations of Table 2 as cal- culated analytically and measured via simulation; again, loops are unrolled to a depth of 4. Note that in all cases analytic and simulation results differ by less than 1%, validating the accuracy of the performance model.... ..."

"... In PAGE 5: ... First we carried out an analysis of the descriptive sta- tistics of the data. We obtained the results shown in the box-plot of figure 1 and eliminated the extreme and atypical data , obtaining the results displayed in Table2 . In this table, we show the descriptive statistics of the valid data for the diagrams that used composite states (CS) and of those that did not.... In PAGE 7: ... 2. Understandability efficiency profile plot from the first experiment Combining the results obtained in Table2 and figure 2, we can reject the hypothe- sis H0, which asserted that the use of composite states did not improve the under- standability efficiency of an UML statechart diagram. 3.... ..."