[R] meta-analysis normal quantile plot metafor

[Viechtbauer Wolfgang (STAT)]

I do not know how Metawin calculates the confidence interval for the QQ-plot. The manual lists Chambers, Cleveland, Kleiner, and Tukey (1983) as a reference for the method used. Not sure whether that is the "classical confidence interval".

The method used in metafor is based on Cook and Weisberg (1982) and you can examine the code to see exactly what is done.

I would start by first checking whether the results from the fixed/random-effects model are essentially the same in Metawin and metafor. Metawin uses the DL-estimator for heterogeneity, so you would want to use that as well in metafor for the random-effects model. Also note that Metawin calculates confidence intervals for the fixed effects (i.e., the mean effect size in the fixed/random-effects model) based on a t-distribution with k-1 degrees of freedom, so you may get a slight difference there (in metafor, the default is to use a normal distribution; if you set knha=TRUE, you will also get a t-distribution with k-1 degrees of freedom, but the additional Knapp & Hartung adjustment to the standard error will be done, so these results are also not directly comparable). But these are minor differences. If the results are not close, then there is some prior difference (before getting the QQ-plot) that needs to be resolved first before the two plots can be compared.

Best,

Wolfgang

--
Wolfgang Viechtbauer, Ph.D., Statistician
Department of Psychiatry and Psychology
School for Mental Health and Neuroscience
Faculty of Health, Medicine, and Life Sciences
Maastricht University, P.O. Box 616 (VIJV1)
6200 MD Maastricht, The Netherlands
+31 (43) 368-5248 | http://www.wvbauer.com
________________________________________
From: r-help-bounces at r-project.org [r-help-bounces at r-project.org] On Behalf Of Ricc [alltheweb2cc+rhelp at gmail.com]
Sent: Monday, January 16, 2012 1:15 PM
To: Michael Dewey
Cc: r-help at r-project.org
Subject: Re: [R] meta-analysis normal quantile plot metafor

Hello,
I used the default parameters:
- envelope: default is TRUE
- level: the default is to take the value from the object (I do not
understand this very well)
- bonferroni : no
- reps : default is 1000
- smooth: default is TRUE
- bass : default is 0

 I used no other arguments.


2012/1/13 Michael Dewey <info at aghmed.fsnet.co.uk>:
> At 15:53 11/01/2012, Ricc wrote:
>>
>> Hello,
>>
>> I once used the metawin software to perform a meta-analysis (see
>> metawinsoft, Rosenberg et al.) and produced normal qqplot to test for
>> a potential bias in the dataset.
>> I now want to re-use the same dataset with the package metafor by W.
>> Viechtbauer (great package btw).
>>
>> I run the qqnorm.rma.uni function. I use standardized effect sizes as
>> in metawin.
>
>
> I think it would help if you said which parameters you used to control the
> envelope. Did you smooth it? Did you use the Bonferroni correction?
>
>
>
>> QQplot generated with metafor differs from the plot obtained with
>> metawin: most of the datapoint fall outside the confidence envelope
>> (using the same confidence level). I don't understand very well how
>> the "pseudo confidence envelope" was created in metafor. Is it more
>> conservative than that from metawin or created using the package
>> envelope ? Unfortunately I do not have access to metawin's code so
>> that I cannot compare implementations but the manual let me think that
>> metawin print classical confidence interval...
>>
>> Thanks for input !
>> Ricc
>>
>> More precisions:
>> R version 2.13.1 (2011-07-08)
>> Platform: x86_64-pc-linux-gnu (64-bit)
>> metafor_1.6-0
>
>
> Michael Dewey
> info at aghmed.fsnet.co.uk
> http://www.aghmed.fsnet.co.uk/home.html
>

______________________________________________
R-help at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.