THE INTEGRATION OF BANKING AND TELECOMMUNICATIONS: THE NEED FOR REGULATORY REFORM

BEST PRACTICES 353

Author 16 23 80 90 91 96 97 98 99 168
16 X 100 94 93 92 94 92 80 93 93
23 100 X 93 93 91 93 91 67 83 92
80 94 93 X 72 75 65 81 86 71 65
90 93 93 72 X 64 66 86 75 80 47
91 92 91 75 64 X 50 58 90 54 62
96 94 93 65 66 50 X 75 86 70 77
97 92 91 81 86 58 75 X 80 85 85
98 80 67 86 75 90 86 80 X 82 91
99 93 83 71 80 54 70 85 82 X 86
168 93 92 65 47 62 77 85 91 86 X
Author
Average

92 89 78 75 71 75 81 81 78 78

Table 3: Cross-Validation Accuracy Scores for Three Edge-Punctuation
Variables

Experiment 2: Modifying the LDFA

By running the LDFA in forward stepwise mode, using
Mahalanobis distance and setting F to enter at 1.84 and F to
remove at 0.71 (SPSS defaults), the accuracy scores improve,
over all ten authors, to 85.9%, as shown in Table 4. In Pair
91/96, none of the variables met the F levels for entering and so
no analysis was run (noted as “nqv” in the table, for “no
qualifying variables”). In the average for this author-pair, the
sums are divided by 8 for the eight comparisons that were
possible (rather than 9).