[R-meta] Question on Forest Plot by subgroups

[Viechtbauer, Wolfgang (NP)]

Dear Gabriel,

Glad to hear you figured things out and nice to report this back to the list.

Best,
Wolfgang

>-----Original Message-----
>From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces using r-project.org] On
>Behalf Of Gabriel Cotlier via R-sig-meta-analysis
>Sent: Tuesday, 20 June, 2023 9:04
>To: R Special Interest Group for Meta-Analysis
>Cc: Gabriel Cotlier
>Subject: Re: [R-meta] Question on Forest Plot by subgroups
>
>Hello all,
>
>Yesterday I posted a solution for the plotting forest plot by subgroups for
>my particular case and I was actually making the mistake when choosing the
>variable for the argument "order" which I corrected in the previously
>posted code (found my own mistake in the argument "order").
>
>But I also posted that was redundant information in my case to use
>subset=(alloc
>== 1) for the subgroups models: res.s, res.r and res.a and indeed it is if
>I use "alloc" as follows:
>
>res.s <- rma(yi, subset=(alloc == 1), vi, data=dat)
>res.r <- rma(yi, subset=(alloc == 1), vi, data=dat)
>res.a <- rma(yi, vi, subset=(alloc == 1),  data=dat)
>
>However, I found out that this gives the same overall result for each
>sub-group as for the total data.
>Therefore I correct it herein as follows:  to get at the end of each
>subgroup its particular overall results the use subset= c() is needed.
>It  worked for me as follows:
>
>### fit random-effects model in the three subgroups
>res.s <- rma(yi, vi, subset= c(Type == "Rainfed"), data=dat)
>res.r <- rma(yi, vi, subset = c(Type == "Experimental"), data=dat)
>res.a <- rma(yi, vi, subset = c(Type == "Computational simulation"),
>data=dat)
>
>In this way it will give the overall model results for each subgroup
>correctly together with the overall result for the complete dataset at the
>end.
>
>Sorry for the confusion.
>Kind regards,
>Gabriel
>
>On Mon, Jun 19, 2023 at 5:49 PM Gabriel Cotlier <gabiklm01 using gmail.com> wrote:
>
>> Hello all,
>>
>> I found the solution to the problem to the code I plotted previously. The
>> problem was that I was ordering by "alloc" which is a column of sing() of
>> correlations and not by the column that divide the groups (3 groups
>> expressed in a column named "Type"  : Computational
>> simulation","Experimental","Natural Rainfed") which in my case is named
>> "Type",after correcting this I also remove unnecessary redundant
>> information subset=(alloc == 1) in each of the models res.s, res.r and
>> res.a. Please find the corrected working code below with highlighted
>> locations where modifications were made  :
>>
>> ############################## CODE
>> ###########################################
>>
>> pdf(file="ForestPlotCompleteDataSetRahul_Test.pdf", width = 12, height =
>> 35)
>>
>> mlabfun <- function(text, res) {
>>   list(bquote(paste(.(text),
>>                     " (Q = ", .(formatC(res$QE, digits=2, format="f")),
>>                     ", df = ", .(res$k - res$p),
>>                     ", p ", .(metafor:::.pval(res$QEp, digits=2,
>> showeq=TRUE, sep=" ")), "; ",
>>                     I^2, " = ", .(formatC(res$I2, digits=1, format="f")),
>> "%, ",
>>                     tau^2, " = ", .(formatC(res$tau2, digits=2,
>> format="f")), ")")))}
>>
>> forest(res,
>>        top = 2,
>>        #xlim=c(-8, 6),
>>        #alim = c(-3.36077, 5.181815),
>>        alim = c(-3.8, 6),
>>        #at=log(c(0.05, 0.25, 1, 4)),
>>        #atransf=exp,
>>        #width = 4,
>>        showweights = TRUE,
>>        ilab=cbind(ni),  # content of extra columns of information
>>        ilab.xpos=c(-4.6), # position of extra columns of information
>>        cex=0.75,
>>        ylim=c(-1, 161), # rows in the y-axis from min-row to max-row with
>> text.
>>        slab = paste(dat$Authors, dat$Year, sep = ", "),
>>        order=Type,
>>        rows=c(2:15,19:85,89:156), # this divide by groups the studies (3
>> groups)
>>        mlab=mlabfun("RE Model for All Studies", res),
>>        #psize= 1 ,
>>        efac = 0.2, # size of the polygon (the diamond)
>>        # header="Author(s) and Year")
>> )
>> # abline(h=res$k+8, lwd=2, col="white")
>>
>> op <- par(cex=0.75, font=2)
>>
>> ### add additional column headings to the plot
>> text(c(-15), res$k+10,  c("Author(s) and Year"), cex = 1.2, font = 2)
>> text(c(-4.6), res$k+10,  c("Samples size"), cex = 1.2, font = 2)
>> text(c(9.7),  res$k+10,  c("Weight"),cex = 1.2, font=2)
>> text(c(9.7),  res$k+9.3, c("%"),cex = 1.2, font=2)
>> text(c(12.6),  res$k+10,  c("Fisher's zr [95% CI]"),cex = 1.2, font=2)
>> text(c(res$k+13), c("Complete Dataset"), cex = 2, font = 2)
>>
>> par(font=4)
>>
>> ### add text for the subgroups
>> text(-17.2, c(16,86,157), pos=4, c("Computational simulation",
>>                                    "Experimental",
>>                                    "Natural Rainfed"), cex = 1.3)
>>
>> par(op)
>>
>> ### fit random-effects model in the three subgroups
>> res.s <- rma(yi, vi, data=dat)
>> res.r <- rma(yi, vi, data=dat)
>> res.a <- rma(yi, vi,  data=dat)
>>
>> ### add summary polygons for the three subgroups
>> addpoly(res.s, row= 88, mlab=mlabfun("RE Model for Subgroup", res.s),efac
>> = 0.2, col='red')
>> addpoly(res.r, row= 18, mlab=mlabfun("RE Model for Subgroup", res.r),efac
>> = 0.2, col='green')
>> addpoly(res.a, row= 1, mlab=mlabfun("RE Model for Subgroup", res.a),efac =
>> 0.2, col='blue')
>>
>>
>> dev.off()
>>
>>
>>
>>
>>
>> On Mon, Jun 19, 2023 at 3:46 PM Gabriel Cotlier <gabiklm01 using gmail.com>
>> wrote:
>>
>>> Hello all,
>>>
>>> I have taken an example from here
>>> <https://www.metafor-project.org/doku.php/plots:forest_plot_with_subgroups>
>and
>>> modified it to my case study of Fisher's z correlations meta-analysis with
>>> the code below.
>>> I would like to plot a forest plot in which each of the subgroups ( 3
>>> groups expressed in a column named "Type"  : Computational
>>> simulation","Experimental","Natural Rainfed") each showing the studies
>>> (rows) with positive correlations and those with negative correlations
>>> divided in agreement to the line of no effect located at zero. I have used
>>> a column named "alloc" which has a 1 for those correlations (rows) with
>>> positive value and -1 for those with negative value, however for some
>>> reason it does not work as expected. It mixes the lines from
>>> different groups, ploting them where they do not correspond and also does
>>> not plot the model summary for the second and third group.
>>> I tried to change several arguments to solve it, however unsuccessfully.
>>>
>>> What could be the error?
>>> Thanks a lot for your help.
>>>
>>> ############################## CODE
>>> ###########################################
>>>
>>> pdf(file="ForestPlotCompleteDataSetbyGroups.pdf", width = 12, height = 35)
>>>
>>> mlabfun <- function(text, res) {
>>>   list(bquote(paste(.(text),
>>>                     " (Q = ", .(formatC(res$QE, digits=2, format="f")),
>>>                     ", df = ", .(res$k - res$p),
>>>                     ", p ", .(metafor:::.pval(res$QEp, digits=2,
>>> showeq=TRUE, sep=" ")), "; ",
>>>                     I^2, " = ", .(formatC(res$I2, digits=1, format="f")),
>>> "%, ",
>>>                     tau^2, " = ", .(formatC(res$tau2, digits=2,
>>> format="f")), ")")))}
>>> forest(res,
>>>        top = 2,
>>>        # xlim=c(-8, 6),
>>>        # alim = c(-3.36077, 5.181815),
>>>        alim = c(-3.8, 6),
>>>        # at=log(c(0.05, 0.25, 1, 4)),
>>>        # atransf=exp,
>>>        # width = 4,
>>>        showweights = TRUE,
>>>        ilab=cbind(ni),  # content of extra columns of information
>>>        ilab.xpos=c(-4.6), # position of extra columns of information
>>>        cex=0.75,
>>>        ylim=c(-1, 161), # rows in the y-axis from min-row to max-row with
>>> text.
>>>        slab = paste(dat$Authors, dat$Year, sep = ", "),
>>>        order=alloc,
>>>        rows=c(2:15,19:85,89:156), # this divide by groups the studies (3
>>> groups) from column named "Type"
>>>        mlab=mlabfun("RE Model for All Studies", res),
>>>        #psize= 1 ,
>>>        efac = 0.2, # size of the polygon (the diamond)
>>>        # header="Author(s) and Year")
>>>        )
>>>
>>> op <- par(cex=0.75, font=2)
>>>
>>> text(c(-15), res$k+10,  c("Author(s) and Year"), cex = 1.2, font = 2)
>>> text(c(-4.6), res$k+10,  c("Samples size"), cex = 1.2, font = 2)
>>> text(c(9.7),  res$k+10,  c("Weight"),cex = 1.2, font=2)
>>> text(c(9.7),  res$k+9.3, c("%"),cex = 1.2, font=2)
>>> text(c(12.6),  res$k+10,  c("Fisher's zr [95% CI]"),cex = 1.2, font=2)
>>> text(c(res$k+13), c("Complete Dataset"), cex = 2, font = 2)
>>>
>>> par(font=4)
>>>
>>> text(-17.2, c(16,86,157), pos=4, c("Computational simulation",
>>>                                  "Experimental",
>>>                                  "Natural Rainfed"), cex = 1.3)
>>> par(op)
>>>
>>> ### fit random-effects model in the three subgroups
>>> res.s <- rma(yi, vi, subset=(alloc == 1), data=dat)
>>> res.r <- rma(yi, vi, subset=(alloc == 2), data=dat)
>>> res.a <- rma(yi, vi, subset=(alloc == 3),  data=dat)
>>>
>>> ### add summary polygons for the three subgroups
>>> addpoly(res.s, row= 88, mlab=mlabfun("RE Model for Subgroup", res.s),efac
>>> = 0.2, col='red')
>>> addpoly(res.r, row= 18, mlab=mlabfun("RE Model for Subgroup", res.r),efac
>>> = 0.2, col='green')
>>> addpoly(res.a, row= 1, mlab=mlabfun("RE Model for Subgroup", res.a),efac
>>> = 0.2, col='blue')
>>>
>>> dev.off()