Skip to main content

Table 5 Wald-type chi-squared tests for inconsistency using detachment factors [D k .T; k (1,…,15)]

From: Network-meta analysis made easy: detection of inconsistency using factorial analysis-of-variance models

Design

Design no. ( k)

Number of studies

Degrees of freedom for D k.T

Effect D k.G.S.T fixed

Effect D k.G.S.T random

D k.T

D k.G.T

D k.T

D k.G.T

Wald statistic

p-value

Wald statistic

p-value

Wald statistic

p-value§

Wald statistic

p-value§

acar:plac

1

1

1

0.09

0.7699

22.45

0.0010

0.02

0.8889

2.25

0.8782

acar:SUal

2

1

1

0.01

0.9091

22.52

0.0010

0.01

0.9430

2.26

0.8765

metf:plac

4

3

1

0.46

0.4976

22.07

0.0012

0.04

0.8379

2.22

0.8814

metf:acar:plac

5

1

2

0.15

0.9297

22.39

0.0004

0.07

0.9634

2.18

0.8129

metf:SUal

6

1

1

15.02

0.0001

7.52

0.2758

1.63

0.2343

0.92

0.9835

piog:plac

8

1

1

5.28

0.0215

17.25

0.0084

0.43

0.5299

1.96

0.9062

piog:metf

9

1

1

5.40

0.0201

17.13

0.0088

0.43

0.5318

1.94

0.9081

piog:rosi

10

1

1

0.05

0.8280

22.49

0.0010

0.01

0.9065

2.27

0.8751

rosi:plac

11

6

1

6.24

0.0125

16.30

0.0122

0.74

0.4112

1.87

0.9168

rosi:metf

12

2

1

0.01

0.9199

22.52

0.0010

0.01

0.9276

2.25

0.8795

rosi:SUal

13

1

1

15.76

<0.0001

6.77

0.3424

1.79

0.2146

0.66

0.9930

  1. Tests are reported for eleven detached designs in the diabetes example of Senn et al. [7] (due to the network structure, the other four designs did not contribute to the inconsistency). The effect Dk.G.S.T was taken either as fixed or as random. Tests significant at the 5% level are boldfaced.
  2. §Based on Wald-type F-test with denominator d.f. computed by Kenward-Roger method [17].