[R-meta] Non-linear model with metafor
Viechtbauer Wolfgang (SP)
wolfgang.viechtbauer at maastrichtuniversity.nl
Sun Sep 24 16:33:48 CEST 2017
I am having a hard time understanding what exactly it is that you did. Would you be able to provide a fully reproducible example (you can also use some mock data) to show your steps?
From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces at r-project.org] On Behalf Of Mustafa Kilickap
Sent: Saturday, 23 September, 2017 11:46
To: r-sig-meta-analysis at r-project.org
Subject: [R-meta] Non-linear model with metafor
Dear metafor users,
I am doing a meta-analysis of epidemiological studies assessing prevalence of hypertension and mean systolic blood pressure (sbp). The data showed that both the prevalence and mean blood pressure decreased over 10 years. The linear models of meta-regression were significant, but residual heterogeneities were also significant. I wanted to assess a non-linear model (y=beta0*t^ (- beta1)) with metafor.
For the prevalence data, I did a meta-regression using the “measure =”PLN” on “log(year)”. The model showed I-squared =0, R-squared =100%. Then, in order to show the non-linear relationship in an understandable way, first I defined the non-linear formula by using the beta parameters estimated in this model, then I did a meta-regression on “year” (measure=PR), using the non-linear formula (mods).
For sbp data, I tried to do the same thing; in order to find the beta0 and beta1 parameters, I intended to model log(sbp) on log(t) with metafor. However, I couldn’t do the analysis as escalc function requires standard deviation (SD) of log transformed sbp, which is difficult to estimate.
I would like to know whether 1) what I did for prevalence data is correct; 2) for sbp data, whether it might be acceptable to do lm(log(sbp)~log(t)) to obtain beta parameters, then test the non-linear function in metaphor; or 3) which method is recommended.
Any help is appreciated
More information about the R-sig-meta-analysis