[R-meta] Forest plot and sub-networks in netmeta

[Gerta Rücker]

Dear Carla,

The help file of the function forest.meta shows that the size of these
boxes (there called squares) is proportional to the study weights (that
may be inverse variances or may be determined otherwise, see argument
weight.study) or (for netmeta) comparison weights. The width of these
squares has no precise meaning, they can be interpreted only relatively
to each other: A study / comparison with greater weight obtains a bigger
square:

weight.study: A character string indicating weighting used to determine
          size of squares or diamonds (argument ‘type.study’) to plot
          individual study results. One of missing, ‘"same"’,
          ‘"fixed"’, or ‘"random"’, can be abbreviated. Plot symbols
          have the same size for all studies or represent study weights
          from fixed effect or random effects model.

The colours of the square have merely practical meaning: The default
colour of the square is gray, and the default colour of the confidence
interval is black. If the study has large weight, the square may become
wider than the confidence interval, and then the confidence interval
(the "cross") becomes white. The square can also be so small that it
looks like a vertical line.

All these colours and properties may be changed by corresponding
arguments, such as col.square, col.square.lines, squaresize. See
help(forest.meta).

Best,
Gerta


On 07/28/2017 05:08 AM, Carla Gomez Creutzberg wrote:
> Greetings everyone,
>
> I am conducting my first network meta-analysis and for that I have
> been using netmeta.
>
> I've read a bit on general and network meta-analysis I can;t seem to
> find any indication as to how to interpret the grey boxes or interval
> demarcations that show up in netmeta's forest plots?
>
> Inline image 1
>
> As shown in the image above (can also be found in the attachments as
> "forest.pl.jpeg"), these boxes appear to be plotted around some (or
> perhaps all) of the model estimates. In some cases they also appear to
> encompass an interval longer than the 95%CI intervals for the estimate
> which is then shown in white instead of black ink. I was wondering
> what type of interval was represented by these boxes and it was
> something that was estimated for all treatments or only some of them?
>
> In addition, for a couple of the meta-analyses I am doing I have found
> that the evidence networks are not well connected and end up as two
> separate sub-networks. I was wondering whether it was appropriate to
> conduct separate meta-analyses on the individual sub-networks or
> whether there was any other way to try and tackle the analysis in
> these cases?
>
> Thanks a lot for attention and any suggestions you can provide
>
> Carla
>
> On Fri, Jul 28, 2017 at 3:07 PM, Carla Gomez Creutzberg
> <cgomezcre at gmail.com <mailto:cgomezcre at gmail.com>> wrote:
>
>     Greetings everyone,
>
>     I am conducting my first network meta-analysis and for that I have
>     been using netmeta.
>
>     I've read a bit on general and network meta-analysis I can;t seem
>     to find any indication as to how to interpret the grey boxes or
>     interval demarcations that show up in netmeta's forest plots?
>
>     Inline image 2
>
>     As shown in the image above (can also be found in the attachments
>     as "forest.pl.jpeg"), these boxes appear to be plotted around some
>     (or perhaps all) of the model estimates. In some cases they also
>     appear to encompass an interval longer than the 95%CI intervals
>     for the estimate which is then shown in white instead of black
>     ink. I was wondering what type of interval was represented by
>     these boxes and it was something that was estimated for all
>     treatments or only some of them?
>
>     In addition, for a couple of the meta-analyses I am doing I have
>     found that the evidence networks are not well connected and end up
>     as two separate sub-networks. I was wondering whether it was
>     appropriate to conduct separate meta-analyses on the individual
>     sub-networks or whether there was any other way to try and tackle
>     the analysis in these cases?
>
>     Thanks a lot for attention and any suggestions you can provide
>
>     Carla
>      
>
>
>
>
> -- 
> *Carla Gómez Creutzberg*
> PhD. Candidate - Tylianakis Lab
> <http://www.tylianakislab.org/the-group.html>
> University of Canterbury - /Te Whare Wānanga o Waitaha//
> /Christchurch, New Zealand
> cgomezcre at gmail.com <mailto:cgomezcre at gmail.com>
>
>
> _______________________________________________
> R-sig-meta-analysis mailing list
> R-sig-meta-analysis at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis


-- 

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

Medical Faculty and Medical Center - University of Freiburg
Institute for Medical Biometry and Statistics

Stefan-Meier-Strasse 26, D-79104 Freiburg, Germany

Phone +49 (0)761 2036673
Fax   +49 (0)761 2036680

Mail  ruecker at imbi.uni-freiburg.de
Web   www.imbi.uni-freiburg.de/biom/

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://stat.ethz.ch/pipermail/r-sig-meta-analysis/attachments/20170728/81e445a8/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: forest.pl.jpg
Type: image/jpeg
Size: 24498 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/r-sig-meta-analysis/attachments/20170728/81e445a8/attachment-0001.jpg>