[R-meta] Ratio of Means
Nathan Pace
n.l.pace at utah.edu
Thu Jul 20 00:05:32 CEST 2017
The outcomes in my example are average drug doses under an experimental and a control condition, so ratio scale measurements.
In a RoM meta analysis there is very high heterogeneity (I2 = 94%).
There is interest in whether a control group mean value as a moderator will improve the model and reduce the heterogeneity.
Adding this moderator does not improve the model or reduce heterogeneity.
It is generally recommended to not use a control group value as a moderator in meta regression of the mean difference (various papers by Sharp, etc).
Does this recommendation still apply when the study values have been transformed into the RoM and not using the mean difference?
Does this require a different model or a Bayesian meta regression?
Your insights are much appreciated.
Nathan
Subject: RE: Ratio of Means
I would say the canonical reference on the log transformed ratio of means (often called the 'response ratio' in the field of ecology) is:
Hedges, L. V., Gurevitch, J., & Curtis, P. S. (1999). The meta-analysis of response ratios in experimental ecology. Ecology, 80(4), 1150-1156.
If you are using metafor, the data used in that paper can be found under dat.curtis1998 (i.e., help(dat.curtis1998) for more details).
This measure is used quite a lot in meta-analyses in that field since the measurements in those studies tend to be ratio scale measurements (e.g., the weight/mass/height of plants grown under different experimental conditions). And in fact, that is the big caveat: This outcome measure is only applicable for ratio scale measurements. Note that the examples in Friedrich, Adhikari, & Beyene (2008) are also all based on things that could be considered ratio scale measurements (urine output, serum creatinine, and creatinine clearance).
So, as long as you are dealing with such measurements, I think the response ratio is a perfectly fine outcome measure for a meta-analysis.
Minor sidenote: In rma.uni(), you should use test="knha" instead of knha=TRUE. The latter still works, but is now undocumented.
Best,
Wolfgang
--
Wolfgang Viechtbauer, Ph.D., Statistician | Department of Psychiatry and
Neuropsychology | Maastricht University | P.O. Box 616 (VIJV1) | 6200 MD
Maastricht, The Netherlands | +31 (43) 388-4170 | http://www.wvbauer.com
-----Original Message-----
From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces at r-project.org] On Behalf Of Nathan Pace
Sent: Wednesday, July 19, 2017 07:19
To: r-sig-meta-analysis at r-project.org
Subject: [R-meta] Ratio of Means
Hi All,
In two papers (J Clin Epidemiol 2011;64:556–564. BMC Med Res Method 2008;8(32)DOI: 10.1186/1471-2288-8-32)
Friedrich, Adhikari, and Beyene proposed methods for the meta analysis of the Ratio of Means of continuous outcomes.
Using the reported means, standard deviations, and sample sizes of the experimental and control groups, the log of the ratio of means and the associated standard error using a first order delta methods are estimated. This meta analysis then uses the generic inverse variance method.
I have a k = 23 meta analysis. I used the metafor function rma.uni with REML and knha = T; a moderator was included.
Any comments?
Any unexpected pitfalls?
Nathan
More information about the R-sig-meta-analysis
mailing list