Intelligent systems in accounting, finance and managementassessing predictive performance of ann-based classifiers, страница 20

I                                              J                                (IJ)                                                                                            for difference

Lower bound

Upper bound

1

2

(5.438)b

0.017

0.000

(5.473)

(5.404)

3

(3.933)b

0.017

0.000

(3.967)

(3.898)

2

1

5.438b

0.017

0.000

5.404

5.473

3

1.506b

0.017

0.000

1.471

1.540

3

1

3.933b

0.017

0.000

3.898

3.967

2

(1.506)b

0.017

0.000

(1.540)

(1.471)

a

1: ‘no preprocessing’; 2: ‘normalization’; 3: ‘maximum of absolute values’.

b

The mean difference is significant at the 0.05 level.

Table XVI. Pairs’ comparison for ‘distribution’ factor

DISTRIBa                            DISTRIBa                       Mean difference                 SE                  Sig.                          95% confidence interval

I                                            J                                (IJ)                                                                                            for difference

Lower bound

Upper bound

1

2

(0.439)b

0.022

0.000

(0.483)

(0.394)

3

0.251b

0.022

0.000

0.206

0.296

4

0.775b

0.022

0.000

0.730

0.819

5

2.984b

0.022

0.000

2.939

3.3028

2

1

0.439b

0.022

0.000

0.394

0.483

3

0.690b

0.022

0.000

0.645

0.734

4

1.213b

0.022

0.000

1.169

1.258

5

3.422b

0.022

0.000

3.378

3.467

3

1

(0.251)b

0.022

0.000

(0.296)

(0.206)

2

(0.690)b

0.022

0.000

(0.734)

(0.645)

4

0.524b

0.022

0.000

0.479

0.568

5

2.733b

0.022

0.000

2.688

2.777

4

1

(0.775)b

0.022

0.000

(0.819)

(0.730)

2

(1.213)b

0.022

0.000

(1.258)

(1.169)

3

(0.524)b

0.022

0.000

(0.568)

(0.479)

5

2.209b

0.022

0.000

2.164

2.254

5

1

(2.984)b

0.022

0.000

(3.028)

(2.939)

2

(3.422)b

0.022

0.000

(3.467)

(3.378)

3

(2.733)b

0.022

0.000

(2.777)

(2.688)

4

(2.209)b

0.022

0.000

(2.254)

(2.164)