ggplot doesn´t order levels since the last update - r

Question

I´m trying to produce this plot here:
But the order on the right side gets mixed up.
In each horizontal line there are two stacked barplots, one in positive, one in negative direction. Each one has its own dataframe. df1 for the left side, df2 for the right side. The middle-category is split in half and partly on the left and the other half on the right side.
I tried to reorder the factor df2$level, which holds the order for the right side barplot, but it didn´t change a thing (of course i took out the order= as.numeric(level) from the ggplot2 call).
df2$level <- factor(df2$level, levels=rev(levels(df2$level)))
df2$level
Here is the example-data:
library("plyr")
library("dplyr")
library("stringr")
library("ggplot2")
# example data
Variable<-c("1","1","1","1","1","2","2","2","2","2","3","3","3","3","3","4","4","4","4","4")
level<-c(5,4,3,2,1,5,4,3,2,1,5,4,3,2,1,5,4,3,2,1)
perc_w<-c(3.70,11.80,10.10,25.80,38.60,2.00,16.90,13.25,28.80,25.80,1.80,6.50,9.35,33.60,39.40,3.50,12.40,14.10,34.80,21.10)
df<-data.frame(Variable,level,perc_w)
df$perc_w<-as.numeric(df$perc_w)
df$level<-as.factor(df$level)
# item text
items<-c("~ It´s not known, if climate change is real",
"~ In my opinion, the risks of climate change are exaggerated by activists",
"~ Climate change is not as dangerous as it is claimed",
"~ I´m convinced that we can handle climate change")
df$Variable<-as.character(df$Variable)
df$Variable[df$Variable==1]<-items[1]
df$Variable[df$Variable==2]<-items[2]
df$Variable[df$Variable==3]<-items[3]
df$Variable[df$Variable==4]<-items[4]
df$Variable<-as.ordered(df$Variable)
# calculate halves of the neutral category
df.split <-df %>% filter(level==3) %>% mutate(perc_w=as.numeric(perc_w/2))
# replace old neutral-category
df<-df %>% filter(!level==3)
df<-full_join(df,df.split) %>% arrange(level) %>% arrange(desc(Variable))
#split dataframe
df1<-df %>% filter(level == 3 | level== 2 | level==1)
df2<-df %>% filter(level == 5 | level== 4 | level==3) %>% mutate(perc_w = perc_w *-1)
# automatic line break
df1$Variable <-str_wrap(df1$Variable, width = 41)
df2$Variable <-str_wrap(df2$Variable, width = 41)
# reorder factor "Variable"
df1$Variable <- factor(df1$Variable, levels=rev(unique(df1$Variable)))
df2$Variable <- factor(df2$Variable, levels=rev(unique(df2$Variable)))
#Plot
p<-ggplot() +
geom_bar(data=df1, aes(x = Variable, y=perc_w, fill = level, order = -as.numeric(level)),position="stack", stat="identity") +
geom_bar(data=df2, aes(x = Variable, y=perc_w, fill = level, order = as.numeric(level)),position="stack", stat="identity") +
geom_hline(yintercept = 0, color =c("black"))+
theme_bw() +
coord_flip() +
guides(fill=guide_legend(title="",reverse=TRUE)) +
scale_fill_brewer(palette="Blues", name="",labels=c("--","-","0","+","++")) +
labs(title=expression(atop(bold("Attitudes towards climate change"),
atop(italic("Some roughly translated items"),""))),
y="percentages",x="") +
theme(legend.position="top",
axis.ticks = element_blank(),
plot.title = element_text(size=25),
axis.title.y=element_text(size=16),
axis.text.y=element_text(size=13),
axis.title.x=element_text(size=16),
axis.text.x=element_text(size=13),
legend.title=element_text(size=14),
legend.text=element_text(size=12)
)
p

Answer 1

Shamelessly taking a hint from this SO question:
ggplot will plot the stacked bars in the order it encounters them when using stat = "identity".
So, adding
df1 <- df1 %>% group_by(Variable) %>% arrange(desc(level))
df2 <- df2 %>% group_by(Variable) %>% arrange(level)
just before your plot code should give you the desired results.