Skip to main content

Table 8 Pairwise differences of the ten treatment means

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

  benf metf migl piog plac rosi sita SUal vild
acar -0.1045 (0.3659) 0.2870 (0.2504) 0.1085 (0.3280) 0.2880 (0.3054) -0.8414 (0.2384) 0.3924 (0.2526) -0.2714 (0.4165) -0.4238 (0.2568) -0.1414 (0.4159)
benf   0.3915 (0.3153) 0.2130 (0.3575) 0.3925 (0.3492) -0.7369 (0.2776) 0.4968 (0.3038) -0.1669 (0.4401) -0.3194 (0.3622) -0.0369 (0.4395)
metf    -0.1785 (0.2703) 0.0010 (0.2176) -1.1284 (0.1494) 0.1053 (0.1600) -0.5584 (0.3727) -0.7109 (0.2272) -0.4284 (0.3721)
migl     0.1795 (0.3093) -0.9499 (0.2253) 0.2839 (0.2569) -0.3799 (0.4091) -0.5324 (0.3238) -0.2499 (0.4085)
piog      -1.1294 (0.2119) 0.1043 (0.2163) -0.5594 (0.4019) -0.7119 (0.2914) -0.4294 (0.4013)
plac       1.2337 (0.1235) 0.5700 (0.3414) 0.4175 (0.2326) 0.7000 (0.3408)
rosi        -0.6637 (0.3631) -0.8162 (0.2290) -0.5337 (0.3624)
sita         -0.1525 (0.4132) 0.1300 (0.4824)
SUal          0.2825 (0.4126)
  1. Means for the diabetes example of Senn et al. [7] computed from model (2), dropping the design × treatment interaction (G.T) and modelling heterogeneity (G.S.T) as random. Table reports pairwise mean differences (and associated standard errors).