[R] How to differ stats_cor labels by group on a ggplot

Sun Jul 12 08:08:03 CEST 2020

Okay. If you get stuck, you can always roll your own:

psdf<-read.table("sim_data.txt",header=TRUE,stringsAsFactors=FALSE)
library(plotrix)
source("supsubtext.R")
par(cex.axis=1.5)
for(i in 3:6) {
 png(paste0("TestAge_",names(psdf[i]),".png"),width=600)
 plot(psdf$Age,psdf[,i],xlim=c(20,85),ylim=c(-0.1,0.5),
  xlab="Age",ylab=names(psdf)[i],
  axes=FALSE,pch=19,col=c("green","orange")[psdf$AgeGroup+1],
  cex=2)
 fullaxis(1,col="black",lwd=1.5)
 fullaxis(2,col="black",lwd=1.5)
 # fit the entire column
 lmfit<-lm(psdf[,i]~psdf$Age)
 lmpf<-predict(lmfit,se.fit=TRUE)
 uniqage<-sort(unique(psdf$Age))
 cis<-supsmu(psdf$Age,lmpf$se.fit)$y
 lmline<-lmfit$coefficients[1]+uniqage*lmfit$coefficients[2]
 # display the overall confidence band
 polygon(c(uniqage,rev(uniqage)),c(lmline+cis,rev(lmline-cis)),
  border=NA,col="lightgray")
 # now the "young" group
 young<-psdf$AgeGroup == 0
 lmfit<-lm(psdf[young,i]~psdf[young,"Age"])
 lmpf<-predict(lmfit,se.fit=TRUE)
 uniqage<-sort(unique(psdf$Age[young]))
 cis<-supsmu(psdf$Age[young],lmpf$se.fit)$y
 lmline<-lmfit$coefficients[1]+uniqage*lmfit$coefficients[2]
 polygon(c(uniqage,rev(uniqage)),c(lmline+cis,rev(lmline-cis)),
  border=NA,col="#88ff8880")
 # now the "old" group
 old<-psdf$AgeGroup == 1
 lmfit<-lm(psdf[old,i]~psdf[old,"Age"])
 lmpf<-predict(lmfit,se.fit=TRUE)
 uniqage<-sort(unique(psdf$Age[old]))
 cis<-supsmu(psdf$Age[old],lmpf$se.fit)$y
 lmline<-lmfit$coefficients[1]+uniqage*lmfit$coefficients[2]
 polygon(c(uniqage,rev(uniqage)),c(lmline+cis,rev(lmline-cis)),
  border=NA,col="#ffaa8880")
 points(psdf$Age,psdf[,i],pch=19,
  col=c("green","orange")[psdf$AgeGroup+1],cex=2)
 # overall correlation
 corval<-cor.test(psdf$Age,psdf[,i])
 corstr<-paste0("R(overall) = ",round(corval$estimate,2),", p = ",
  round(corval$p.value,2))
 supsubtext(20,0.05,corstr,sub=2:10,col="blue")
 # young correlation
 corval<-cor.test(psdf[young,"Age"],psdf[young,i])
 corstr<-paste0("R(young) = ",round(corval$estimate,2),", p = ",
  round(corval$p.value,2))
 supsubtext(20,0,corstr,sub=2:8,col="green")
 # old correlation
 corval<-cor.test(psdf[old,"Age"],psdf[old,i])
 corstr<-paste0("R(old) = ",round(corval$estimate,2),", p = ",
  round(corval$p.value,2))
 supsubtext(20,-0.05,corstr,sub=2:6,col="orange")
 legend(40,0.5,c("Overall","Young","Old"),pch=19,
  col=c("lightgray","green","orange"))
 dev.off()
}

I've been meaning to finish off the "supsubtext" function for a while.

Jim

On Sun, Jul 12, 2020 at 11:20 AM Paulina Skolasinska
<paulina.skol using hotmail.com> wrote:
>
> Sorry for misspelling your name, Jim. Well, it seems this is not worth the effort then. If my advisor decides this is absolutely essential, I’ll add it in gimp or something. Thanks for giving it a go, Jim.
>
> Get Outlook for iOS
> ________________________________
> From: Jim Lemon <drjimlemon using gmail.com>
> Sent: Saturday, July 11, 2020 6:32:13 PM
> To: Paulina Skolasinska <paulina.skol using hotmail.com>
> Cc: r-help using r-project.org <r-help using r-project.org>
> Subject: Re: [R] How to differ stats_cor labels by group on a ggplot
>
> Hi Paulina,
> Thanks for the data. Even after downloading a ton of packages along
> with the "ggpubr" package, I get a namespace error when I try to load
> it. After a bit of wrangling, I did get the code to run without the
> "stat_cor" function, producing the four PNG images. My best guess is
> to try the geom_label function with stat="cor", but I can't see any
> argument to format the actual label. I think this is about as far as
> I'm going to get.
>
> Jim
>
> On Sun, Jul 12, 2020 at 12:09 AM Paulina Skolasinska
> <paulina.skol using hotmail.com> wrote:
> >
> > Thanks Tim, here is the link to the data: https://we.tl/t-c4x9Lw7LeR
> > I'm posting my code again so it better matches the modified data, and because I've added several improvements in the meantime.
> >
> > for (i in 3:6) {
> >     p1 <- ggplot(df, mapping=aes(x = Age, y = unlist(df[i]), color=factor(AgeGroup))) +
> >         geom_smooth(method="lm") +
> >         geom_point(size = 4) +
> >         scale_x_continuous(breaks = seq(20,90, by=10)) +
> >         scale_y_continuous(breaks = seq(0.1,0.5, by=0.1)) +
> >         theme_classic() +
> >         expand_limits(y = 0.5) +
> >         ylab(names(df)[i]) +
> >         theme(axis.text.x = element_text(face="bold", size=14),
> >               axis.text.y = element_text(face="bold", size=14),
> >               axis.title.x = element_text(size=14, face="bold"),
> >               axis.title.y = element_text(size=14, face="bold"),
> >               legend.title = element_text(color="black", size=12),
> >               legend.text = element_text(colour="black", size = 14, face = "bold")) +
> >         geom_smooth(data=df, mapping = aes(x = Age, y = unlist(df[i]), group=1, color="black"), method = "lm") +
> >         scale_colour_discrete(name="Group", labels=c("Young", "Old", "Overall")) +
> >         stat_cor(aes(color = factor(AgeGroup), y = unlist(df[i])),
> >                  method = "pearson", label.x.npc = c("center"), label.y.npc="top") +
> >         stat_cor(aes(x = Age, y = unlist(df[i]),
> >                  label = paste(sub("R",expression("R"[overall]),..r.label..), ..p.label.., sep = "~`,`~"),
> >                  group=1, color="black"), method = "pearson",label.y=0.5, label.x.npc = c("center"),
> >                  position = position_nudge(y = 0.015 * 0.5))
> >
> >     ggsave(p1, file=paste0("TestAge_", names(df)[i], ".png"), scale=1, width=16, height=10, units="cm")
> > }
> >
> >
> > ________________________________
> > Od: Jim Lemon <drjimlemon using gmail.com>
> > Wysłane: sobota, 11 lipca 2020 04:09
> > Do: Paulina Skolasinska <paulina.skol using hotmail.com>
> > DW: r-help using r-project.org <r-help using r-project.org>
> > Temat: Re: [R] How to differ stats_cor labels by group on a ggplot
> >
> > Hi Paulina,
> > Without data it's hard to work out what you are doing. Even a small
> > simulated data set would help to get answers.
> >
> > Jim
> >
> > On Fri, Jul 10, 2020 at 11:49 PM Paulina Skolasinska
> > <paulina.skol using hotmail.com> wrote:
> > >
> > > 'm using ggplot2 to plot two variables at a time. I'm plotting two age groups and overall data on the same graph. I'm also using stat_cor form the ggpubr package to report correlations for the two groups and overall data.
> > >
> > > I want each stat_cor label to have a custom subscript - the group name ("old", "young"). I have managed to do this for the overall data, but I don't know how to add custom labels for each group separately. I'd like the labels to look like this: https://imgur.com/a/naF7uNW
> > >
> > > > for (i in 18:21) {
> > >   p1 <- ggplot(df, mapping=aes(x = Age, y = unlist(df[i]), color=factor(AgeGroup))) +
> > >     geom_smooth(method="lm") +
> > >     geom_point(size = 4) +
> > >     geom_smooth(data=df, mapping = aes(x = Age, y = unlist(df[i]), group=1, color="black"), method = "lm") +
> > >     scale_colour_discrete(name="Group", labels=c("young", "old", "overall")) +
> > >     stat_cor(aes(color = factor(AgeGroup), y = unlist(df[i]))) +
> > >     stat_cor(aes(x = Age, y = unlist(df[i]), group=1, color="black",
> > >                  label = paste(sub("R",expression("R"[overall]),..r.label..), ..p.label.., sep = "~`,`~")))
> > >
> > >   ggsave(p1, file=paste0("Age_", names(df)[i], ".png"), scale=1)
> > > }
> > > ______________________________________________
> > > R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> > > https://stat.ethz.ch/mailman/listinfo/r-help
> > > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> > > and provide commented, minimal, self-contained, reproducible code.