[R-meta] Inconsistency Testing
N|cky@We|ton @end|ng |rom br|@to|@@c@uk
Tue Mar 9 23:51:09 CET 2021
You need loops of evidence in your network in order to be able to test for inconsistency. I don’t know what your network diagram looks like, but if there are no loops then it isn’t possible to test for inconsistency (but that doesn’t mean it isn’t there!) Note that rather than lump treatments together another option is to fit a "class model" where you have a hierarchical model of treatment within class defined by mechanism of action. Then you can estimate a between treatment within class variability.
From: R-sig-meta-analysis <r-sig-meta-analysis-bounces using r-project.org> On Behalf Of Sean
Sent: 09 March 2021 22:22
To: r-sig-meta-analysis using r-project.org
Subject: [R-meta] Inconsistency Testing
I’ve been stuck on a question about inconsistency testing for quite some time, but first a little simplified background:
I’ve calculated effect sizes for all treatments from 50 independent trials conducted over the past 10 years. These treatments are different fungicides applied to a plant to control a foliar pathogen.
Throughout those 10 years, researchers tested 20 different products, and a treatment (4-15 per trial) is different combinations of usually
1-6 of those fungicides. There was no coordination over those 10 years in experimental design, so no treatment was truly replicated. Instead, what I’ve done is reduce treatments into larger categories based on the modes of action of those fungicides. This has allowed me to have enough similarly coded treatments to perform a network meta-analysis.
That went all well and good, however, when it comes to inconsistency testing, I have as many study designs as I have studies. 50 independent trials, 50 designs.
Can I even technically perform inconsistency testing? What I've read in the literature doesn't seem to account for my situation. If not, what does this mean for my meta-analysis? Do I truly need to perform inconsistency testing?
Thank you all for your time, hope your week is going well!
R-sig-meta-analysis mailing list
R-sig-meta-analysis using r-project.org
More information about the R-sig-meta-analysis