[R-meta] [R=meta] Question about removal of outliers and power calculation

Gerta Ruecker ruecker @end|ng |rom |mb|@un|-|re|burg@de
Tue Jul 30 14:15:24 CEST 2019

Dear Joanne,

I agree with Michael in all points. I had a look at the page you 
referred to and it did not became clear to me why they want to calculate 
power. At least the last caveat on that page is true: "Please note that 
power analyses should always be conducted a priori, meaning before you 
perform the meta-analysis." In my view, it all comes down to the same 
thing (whether you calculate numbers before or afterwards) because, as 
Michael said, you cannot plan the size of a meta-analysis like you can 
plan the size of a study.

Note that the reference you gave is not a guideline, but simply a book, 
a practical guide.



Am 30.07.2019 um 13:38 schrieb Cath Kids:
> Dear Michael,
> Thank you for your reply! Please see more response in blue.
> On Tue, Jul 30, 2019 at 6:46 PM Michael Dewey <lists using dewey.myzen.co.uk>
> wrote:
>> Dear Joanne
>> In line comments
>> On 30/07/2019 04:21, Cath Kids wrote:
>>> Hello everyone,
>>> I am new to meta-analysis and I would like to clarify some conceptual
>>> matter:
>>> 1. Should I remove outliers before doing subgroup analysis/ meta
>>> regression? In my study, heterogeneity became insignificant after removal
>>> outliers. I read meta-analyses which did both practice and I wonder which
>>> is the correct way.
>> This raises a number of issues. If heterogeneity exists why do you want
>> to reduce or eliminate it? Would it not be better to try to describe and
>> explain it? In general removing outliers leads to a model which is data
>> dependent rather than the scientific model you started with. The only
>> time I would contemplate removing observations would be if there was
>> reason to suspect that they are fraudulent, recorded erroneously, did
>> not really meet the inclusion criteria after all, or some other
>> principled explanation which i cannot think of just now.
> I see. So I should use all observations for the subgroup analysis/
> meta-regression. But then given the heterogeneity of the results, I can
> remove the outliers and check whether there is any change in the overall
> effect size as a sensitivity analysis. is that right?
>>> 2. I wonder whether any one of you are familiar with any tools to
>> calculate
>>> power for meta-analysis of correlation coefficients?
>> If you are planning a study why do you need power? The number of primary
>> studies you find in your literature search is not within your powers to
>> choose. You can only analyse what you find, not what you would like to
>> find. If you have already done the study then all the required
>> information about precision is contained within the confidence estimates
>> about your estimate(s).
>> I refer to this guideline online
> https://bookdown.org/MathiasHarrer/Doing_Meta_Analysis_in_R/power-analysis.html
> .
> However, it does not mention meta-analysis of correlation. So I'm not sure
> how to do it.
>>> Thank you very much!
>>> Regards,
>>> Joanne
>>>        [[alternative HTML version deleted]]
>>> _______________________________________________
>>> R-sig-meta-analysis mailing list
>>> R-sig-meta-analysis using r-project.org
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
>>> ---
>>> This email has been checked for viruses by AVG.
>>> https://www.avg.com
>> --
>> Michael
>> http://www.dewey.myzen.co.uk/home.html
> 	[[alternative HTML version deleted]]
> _______________________________________________
> R-sig-meta-analysis mailing list
> R-sig-meta-analysis using r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis


Dr. rer. nat. Gerta Rücker, Dipl.-Math.

Institute of Medical Biometry and Statistics,
Faculty of Medicine and Medical Center - University of Freiburg

Stefan-Meier-Str. 26, D-79104 Freiburg, Germany

Phone:    +49/761/203-6673
Fax:      +49/761/203-6680
Mail:     ruecker using imbi.uni-freiburg.de
Homepage: https://www.imbi.uni-freiburg.de/persons/ruecker/person_view

More information about the R-sig-meta-analysis mailing list