Message-ID: <8d6a8d63-47d1-4382-7e65-017a78d4f5f6@dewey.myzen.co.uk>
Date: 2023-03-19T13:43:50Z
From: Michael Dewey
Subject: [R-meta] Choice of MA
In-Reply-To: <CAOYO_yA-H-vxJdUSaP7XSa5vCxmnkgL8dujxc+fM1t61zfF6PA@mail.gmail.com>
Dear Jorge
I suspect opinions will differ here so my comments (in-line) shouls be
taken with a pinch of salt.
Michael
On 19/03/2023 10:20, Jorge Teixeira via R-sig-meta-analysis wrote:
> Hi all. I am doing a SRMA in which most outcomes have just 4 or 5 studies.
> One of them has 4 arms (3 with the intervention of interest; 1 with CON).
>
>
> Questions:
>
>
> 1) Would you combine these 3 groups into one and do a standard MA or would
> you do a multi-level MA? Note: I have access to the database of the study
> with 4 arms.
Usually I would say do an NMA but with so few studies combining the
intervention groups might be fine as long as they ae scientifically
combinable.
>
> 2) Side question: would you tend to not report funnel plots or Egger?s
> regression in this case?
>
I would not draw a funnel plot or do any of the regression tests.
>
> 3) If I broaden my eligibility criteria and had 2 more studies, would your
> answers differ?
>
Not more 2 more, no.
>
> Thank you,
>
> JT
>
> [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-meta-analysis mailing list @ R-sig-meta-analysis at r-project.org
> To manage your subscription to this mailing list, go to:
> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
>
--
Michael
http://www.dewey.myzen.co.uk/home.html