Skip to main content

Advertisement

Table 9 ANOVA tests

From: The effects of class rarity on the evaluation of supervised healthcare fraud detection models

Term Df Sum Sq Mean Sq F value Pr(>F)
(a) Four factor: Train_Test
Datasets 3 1.96516 0.65505 1600.558663 0
Learner 2 0.66409 0.33205 811.3215437 1.29E−270
Pos count 4 2.33669 0.58417 1427.367028 0
Ratio 5 0.30092 0.06018 147.0525251 1.17E−136
Dataset:learner 6 0.22145 0.03691 90.18221558 2.05E−102
Dataset:pos count 7 0.05788 0.00827 20.2044874 1.61E−26
Learner:pos count 8 0.03291 0.00411 10.0523189 7.04E−14
Dataset:ratio 15 0.08724 0.00582 14.21108373 2.17E−35
Learner:ratio 10 0.46537 0.04654 113.7087576 2.69E−194
Pos count:ratio 20 0.05443 0.00272 6.649302133 3.98E−18
Dataset:learner:pos count 14 0.02401 0.00171 4.189763574 2.52E−07
Dataset:learner:ratio 30 0.11577 0.00386 9.428942232 2.99E−40
Dataset:pos count:ratio 35 0.01306 0.00037 0.911606914 0.61790442
Learner:pos count:ratio 40 0.10682 0.00267 6.52539348 3.70E−32
Dataset:learner:pos count:ratio 70 0.04795 0.00068 1.673573178 4.56E−04
Residuals 2430 0.99452 0.00041
(b) Four factor: Train_CV
Datasets 7 26.98719 3.85531 2641.872038 0
Learner 2 1.34356 0.67178 460.3389767 5.33E−194
Pos count 3 6.30292 2.10097 1439.702651 0
Ratio 5 1.25714 0.25143 172.2919289 3.65E−178
Dataset:learner 14 0.79698 0.05693 39.00963177 2.86E−105
Dataset:pos count 4 0.34344 0.08586 58.83679308 2.58E−49
Learner:pos count 6 0.07181 0.01197 8.201310311 7.04E−09
Dataset:ratio 35 0.32743 0.00936 6.410645904 3.43E−29
Learner:ratio 10 1.36440 0.13644 93.49620948 1.03E−187
Pos count:ratio 15 0.13786 0.00919 6.298014086 1.64E−13
Dataset:learner:pos count 8 0.04741 0.00593 4.061371638 7.71E−05
Dataset:learner:ratio 70 0.42329 0.00605 4.143770726 3.52E−28
Dataset:pos count:ratio 20 0.01735 0.00087 0.594299086 0.919852682
Learner:pos count:ratio 30 0.08452 0.00282 1.9304991 0.001663456
Dataset:learner:pos count:ratio 40 0.07660 0.00192 1.312296909 0.089594623
Residuals 13230 19.30669 0.00146
(c) Two factor: Train_Test and Train_CV
Evaluation method 1 0.45439 0.45439 103.9411081 2.59E−24
Residuals 12598 55.07319 0.00437