Table 5 |
||
|
Summary of the validation with data set 1 |
||
|
MIX |
STATA/CMA * |
|
|
|
||
|
Items in odds ratio meta-analysis |
Results |
Results |
|
|
||
|
Fixed effect (IV) odds ratio |
||
|
OR – IV (95% CI) |
0.7677 (0.7196 to 0.8190) |
0.7677 (0.7196 to 0.8190) |
|
Z (P value) |
8.0073 (< 0.0001) |
8.01 (< 0.0001) |
|
Q (P value) |
39.48 (0.17) |
39.48 (0.17) |
|
I2 (95% CI) |
0.1895 (0 to 0.4749) |
0.1895 |
|
Fail-safe N (tolerance level) |
270 (175) |
270 * |
|
Rank correlation tau ** (P value) |
0.1799 (0.141) |
(0.141) |
|
Egger's regression intercept (95% CI) |
-0.2955 (-0.7880 to 0.1970) |
-0.2955 (-0.7880 to 0.1970) |
|
Macaskill's regression slope (P value) |
0 (0.8396) |
0 (0.840) |
|
Trim-and-fill association measure (95% CI) |
0.7744 (0.7260 to 0.8260) |
0.774 (0.726 to 0.826) |
|
Trim-and-fill imputed studies |
6 |
6 |
|
Random effects (IV + t) odds ratio |
||
|
OR – IV + t (95% CI) |
0.7619 (0.6825 to 0.8506) |
0.7619 (0.6825 to 0.8506) |
|
Z (P value) |
-4.84 (< 0.0001) |
-4.84 (< 0.0001) |
|
t2 |
0.117 |
0.117 |
|
Trim-and-fill association measure (95% CI) |
0.7828 (0.6907 to 0.8871) |
0.783 (0.691 to 0.887) |
|
Trim-and-fill imputed studies |
6 |
6 |
|
|
||
|
This table shows the odds ratio validation of the inverse variance fixed effect and random effects analyses with data set 1. Though not shown here, the results from Mantel-Haenszel and Peto weighting were evaluated similarly and so were the other association measures (risk difference and risk ratio). The process was repeated for each data set. (*): Results marked with '*' were produced by CMA. (**): The rank correlation tau is a continuity corrected tau-b. Abbreviations: "IV" = inverse variance weighting, "OR" = odds ratio, "CI" = confidence interval, "IV+t" = inverse variance plus tau, which refers to random effects weighting according to the DerSimonian-Laird method. |
||
|
Bax et al. BMC Medical Research Methodology 2006 6:50 doi:10.1186/1471-2288-6-50 |
||
