前回の続きで、別のパッケージを教えてもらった。
install.packages("meta") install.packages("rmeta") library(meta) library(rmeta) data(catheter, package="rmeta") a <- catheter[complete.cases(catheter$col.trt, catheter$n.trt, catheter$col.ctrl, catheter$n.ctrl),] b <- metabin(col.trt, n.trt, col.ctrl, n.ctrl, data=a, sm="OR", studlab=Name, comb.fixed=TRUE, comb.random=TRUE) windows(width=7, height=7); par(lwd=1, las=1, family="sans", cex=1) forest.meta(b)
windows(width=7, height=7); par(lwd=1, las=1, family="sans", cex=1) funnel(b)
b
OR 95%-CI %W(fixed) %W(random)
Ciresi 0.6946 [0.3400; 1.4192] 6.10 9.90
George 0.1176 [0.0425; 0.3254] 7.20 6.96
Hannan 0.8261 [0.3978; 1.7153] 5.29 9.71
Heard 0.6031 [0.3838; 0.9477] 16.20 13.14
vanHeerden 0.2667 [0.0712; 0.9991] 2.97 4.96
Maki 0.4898 [0.2924; 0.8206] 14.04 12.32
Bach(a) 0.0651 [0.0031; 1.3643] 1.56 1.23
Ramsay 0.5844 [0.3729; 0.9158] 16.72 13.18
Appavu 0.5455 [0.0287; 10.3683] 0.39 1.31
Trazzera 0.4673 [0.2325; 0.9392] 7.74 10.09
Collins 0.0950 [0.0219; 0.4114] 6.77 4.27
Bach(b) 0.1107 [0.0249; 0.4934] 5.24 4.14
Tennenberg 0.2190 [0.0969; 0.4947] 9.79 8.81
Number of studies combined: k=13
OR 95%-CI z p.value
Fixed effect model 0.4396 [0.3601; 0.5367] -8.0700 < 0.0001
Random effects model 0.3915 [0.2758; 0.5558] -5.2459 < 0.0001
Quantifying heterogeneity:
tau^2 = 0.1900; H = 1.5 [1.1; 2.04]; I^2 = 55.3% [16.7%; 76.1%]
Test of heterogeneity:
Q d.f. p.value
26.87 12 0.0081
Details on meta-analytical method:
- Mantel-Haenszel method
metabias(b) Linear regression test of funnel plot asymmetry (efficient score) data: b t = -2.145, df = 11, p-value = 0.05513 alternative hypothesis: asymmetry in funnel plot sample estimates: bias se.bias slope -1.9353989 0.9022968 -0.1745479
