[R-meta] Meta-analysis for agronomical data

[Emerson Del Ponte]

Dear Norman,

I am a plant pathologist and have used R (mainly metafor package) in
my meta-analytic studies. For the disease-yield relationship, you can
check a reproducible report of the analysis for a paper I used metafor
(escalc function, etc).
https://emdelponte.github.io/paper-white-mold-meta-analysis/
The codes are here:
https://github.com/emdelponte/paper-white-mold-meta-analysis

As to the number of data points in a single trial, it is better if
they are distributed along a good range of disease and yield.  It does
not help much if you have a cluster of several low disease (high
yield) and one high disease (low yield). Usually, fungicide effects
will vary and so a range is expected for these two variables.  You can
reasonably assume that the relationship is linear, but more certain if
you have more trials too. For three diseases I studied, the linear
relationship could be assumed.

Hope this helps.

Emerson

Em ter., 12 de mai. de 2020 às 20:12, Norman DAURELLE
<norman.daurelle using agroparistech.fr> escreveu:
>
> Hello,I would like to use a meta-analysis method to summarize the scientific literature that has been published about the relationship between a plant disease (Blackleg, caused by the fungus Leptosphaeria maculans) and oil rapeseed yield.I have read about meta-analyses and I tried using a meta-analysis method, the one explained by Daniel Quintana in the paper " From pre-registration to publication : a non-technical primer for conducting a meta-analysis to synthesize correlational data" and the youtube video that relates to it, but the object of my interest is more the estimated slope of the regression between disease level and yield than the correlation. Also, the escalc function of the Metafor package requires at least 4 "replications" (or I don't know exactly what to call it, but it is a number of individuals for the sample size of each study in medical sciences) for each effect-size, each study, and in my data what comes closest to a number of individuals would be the number of re
>  plications made for each (fungicide or other) treatment, which is sometimes smaller than 4.to be more explicit I used this line of code :data <- escalc(measure="ZCOR", ri=r, ni=number_of_rep, data=data, slab=paste(author, year, sep=", ")) and I get this error message :
> Warning message:
> In escalc(measure = "ZCOR", ri = r, ni = number_of_rep, data = data,  :
>   Cannot estimate the sampling variance when ni <= 4.I multiplied my "number_of_rep" by 10 to see what results the analysis would give, but I am not confident in the fact that this is a rigorous thing to do.Furthermore I have only seven studies, and from what I read in the plant pathology literature, it seems like a pretty small number of studies. As a comparison, in a paper published in 2009 by Madden et al. (Meta-Analysis for Evidence Synthesis in Plant Pathology : An Overview), they used 101 studies for their meta-analysis... So that made me think maybe there was altogether not enough data for me to use in my meta-analysis, but I would really like to make sure that I have tried everything I could to make it work.Do you have any advice regarding what package and what function to use that would be suitable for a dataset including only 7 studies, and which doesn't have the exact equivalent for the number of individuals that make up the sample sizes ?I would be really grateful.Thank y
>  ou !Norman
>         [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-meta-analysis mailing list
> R-sig-meta-analysis using r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis



-- 
Emerson M. Del Ponte
Universidade Federal de Viçosa, Brazil
Chair of the Graduate Studies in Plant Pathology
EIC for Tropical Plant Pathology
Co-Founder of Open Plant Pathology
My websites: Twitter | GitHub | Google Scholar | ResearchGate
Tel +55 31 36124830