Connecting points jittered by position_jitterdodge

[richardbeare]

Hello,
I am attempting to combine boxplots and points, and connect points together. There are some old discussions saying that the input data needs to be sorted to ensure that geom_line and geom_point receive the same jittering, but this does not seem to work when there is a nested grouping:

library(tidyverse)
set.seed(77)
a1 <- tibble(ID=factor(letters[1:10]), y=runif(10), TP="A", Grp=rep(c("C","P"), 5))
a2 <- tibble(ID=factor(letters[1:10]), y=runif(10), TP="B", Grp=rep(c("C","P"), 5))
a1 <- bind_rows(a1, a2)
a1 <- mutate(a1, TP=factor(TP), Grp=factor(Grp))
a1 <- arrange(a1, ID, TP)

pj <- position_jitter(width=0.2, seed=9)
ggplot(a1, aes(x=TP, y=y, fill=Grp)) +
  geom_boxplot(alpha=0.2, outlier.colour = NA) +
  geom_point(position=pj, aes(colour=Grp, group=ID)) +
  geom_line(position=pj, aes(group = ID, colour=Grp))

This is performing as expected. However the jittering is not separated per group (i.e pink points are not lined up with the pink boxplot). Lets use jitterdodge to achieve that.

pj <- position_jitterdodge(jitter.width=0.4, seed=9)
ggplot(a1, aes(x=TP, y=y, fill=Grp)) +
  geom_boxplot(alpha=0.2, outlier.colour = NA) +
  geom_point(position=pj, aes(colour=Grp, group=Grp))

Now the colours are lined up, attempt to connect together.

ggplot(a1, aes(x=TP, y=y, fill=Grp)) +
  geom_boxplot(alpha=0.2, outlier.colour = NA) +
  geom_point(position=pj, aes(colour=Grp, group=Grp)) +
  geom_line(position=pj, aes(group = ID, colour=Grp), alpha=0.3)

As far as I can tell, none of the following combinations of sorting the input data fix the problem of lines being connected to the points:

a1 <- arrange(a1, ID, TP, Grp)
a1 <- arrange(a1, ID, Grp, TP)
a1 <- arrange(a1, Grp, ID, TP)
a1 <- arrange(a1, Grp, TP, ID)
a1 <- arrange(a1, TP, ID, Grp)
a1 <- arrange(a1, TP, Grp, ID)

[smouksassi]

try a geom that combine point and line see below:

library(tidyverse)
library(lemon)
set.seed(77)
a1 <- bind_rows(tibble(ID=factor(letters[1:10]), y=runif(10), TP="A", Grp=rep(c("C","P"), 5)),
                tibble(ID=factor(letters[1:10]), y=runif(10), TP="B", Grp=rep(c("C","P"), 5))) %>%
  mutate(TP=factor(TP), Grp=factor(Grp)) %>% 
  arrange(ID, TP)

pj <- position_jitterdodge(jitter.width=0.2, seed=9,
                           jitter.height = 0,
                           dodge.width = 0.75)
ggplot(a1, aes(x=TP, y=y, fill=Grp)) +
  geom_boxplot(alpha=0.2, outlier.colour = NA, width =0.75) +
  lemon::geom_pointpath(aes(colour=Grp, group=interaction(ID,Grp)),
                        position = pj)

^{Created on 2020-07-01 by the reprex package (v0.3.0)}

[richardbeare]

@smouksassi
Nice work-around!

Although, having checked a bit deeper, I don't seem to be able to get the jittering to work correctly. Ignoring the lines for the moment, points should appear as below - line up with the boxplots. Currently they are scattered across both...

library(tidyverse)
set.seed(77)
a1 <- tibble(ID=factor(letters[1:10]), y=runif(10), TP="A", Grp=rep(c("C","P"), 5))
a2 <- tibble(ID=factor(letters[1:10]), y=runif(10), TP="B", Grp=rep(c("C","P"), 5))
a1 <- bind_rows(a1, a2)
a1 <- mutate(a1, TP=factor(TP), Grp=factor(Grp))
a1 <- arrange(a1, ID, TP)


pj <- position_jitterdodge(jitter.width=0.2, seed=9,
                           jitter.height = 0,
                           dodge.width = 0.75)
ggplot(a1, aes(x=TP, y=y, fill=Grp)) +
  geom_boxplot(alpha=0.2, outlier.colour = NA, width =0.75) +
  geom_point(aes(colour=Grp, group=interaction(Grp)),
                        position = pj)

[thomasp85]

I think the best way forward is to make geom_pointpath work correctly. There is a lot going on when trying to ensure that jittering is reproducible across layers and I'm not sure we can solve it correctly for all cases

[Syutenjyo]

library(tidyverse)
set.seed(77)

a1 <- bind_rows(tibble(ID=factor(letters[1:15]), y=runif(15), TP="A", Grp=rep(c("C","P","Q"), 5)),
                tibble(ID=factor(letters[1:15]), y=runif(15), TP="B", Grp=rep(c("C","P","Q"), 5)),
                tibble(ID=factor(letters[1:15]), y=runif(15), TP="C", Grp=rep(c("C","P","Q"), 5)),
) %>% mutate(TP=factor(TP), Grp=factor(Grp))


a2 <- a1 %>% arrange(ID, TP, Grp)

pj <- position_jitterdodge(jitter.width=0,
                           jitter.height = 0,
                           dodge.width = 0.75,
                           seed=9)

pp4 <- ggplot(mapping = aes(x=TP, y=y, fill=Grp)) +
  geom_boxplot(data=a1, alpha=0.2, outlier.colour = NA, width =0.75) +
  geom_point(data=a2, aes(colour=Grp, group=interaction(ID,Grp)),
             position = pj) +  
  geom_line(data=a2, aes(colour=Grp, group=interaction(ID,Grp)),
             position = pj)


pp4