[R-meta] Outliers in meta analysis

[Michael Dewey]

Dear Mutlu

I think you first have to decide why you are looking for outliers and 
what you intend to do with them when you find them. Only then will it be 
clear to you which of the methods you need.

Michael

On 11/10/2018 12:34, Mutlu Umaroglu wrote:
> Dear all,
> 
> I am interested in Meta Analysis and studying on outliers. There are 
> several method proposed to examine outliers (Viechtbauer 2010).
> According to metafor packages following studies are outlier in this 
> example. How can we deside exactly which studies are
> influences/outliers?
> 
> Sincerely,
> Mutlu
> 
> 
> library("metafor")
> data <- dat.bcg
> dat1 <- escalc(measure="RD", ai=tpos, bi=tneg, ci=cpos, di=cneg, data=data)
> res1 <- rma(yi, vi, method="DL",data=dat1)
> influence(res1)
> 
>     rstudent  dffits cook.d  cov.r tau2.del   QE.del    hat  weight dfbs 
> inf
> 1   -1.4026 -0.0889 0.0079 0.9976        0 273.7960 0.0031  0.3060 -0.0891
> 2   -3.6289 -0.3913 0.1526 0.9563        0 260.2482 0.0067  0.6656 -0.4008
> 3   -1.7694 -0.1874 0.0351 0.9961        0 271.6713 0.0084  0.8424 -0.1885
> 4   -2.5810 -1.9330 2.4532 0.7580        0 167.7548 0.1190 11.9010 
> -1.9186   *
> 5    1.2615  0.5217 0.2813 1.1661        0 276.3026 0.1143 11.4253 0.5218
> 6   -9.2559 -1.4557 2.0527 0.6846        0 179.5183 0.0113  1.1260 
> -1.7385   *
> 7   -0.8835 -0.2354 0.0551 1.0453        0 271.0926 0.0545  5.4510 -0.2361
> 8    1.6691  1.4621 3.6567 1.8187        0 263.8946 0.1296 12.9579 
> 1.5125   *
> 9    1.1381  0.5412 0.3114 1.2042        0 274.9691 0.1216 12.1644 0.5424
> 10  -3.4172 -0.9551 0.8758 0.9605        0 248.2341 0.0520  5.2024 
> -0.9795   *
> 11   1.3748  1.1033 1.7292 1.5523        0 274.9184 0.1287 12.8677 
> 1.1272   *
> 12   1.8398  0.7986 0.6771 1.2027        0 274.4115 0.1217 12.1716 
> 0.8004   *
> 13   1.6603  1.3631 2.9509 1.7076        0 270.6094 0.1292 12.9187 
> 1.4029   *
> 
> _______________________________________________
> R-sig-meta-analysis mailing list
> R-sig-meta-analysis using r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
> 

-- 
Michael
http://www.dewey.myzen.co.uk/home.html