Bars in geom_bar have unwanted different widths when using facet_wrap

Question

I can'd find a solution for the following problem(s). I would appreciate some help a lot!

The following code produces bar charts using facet. However, due to "extra space" ggplot2 has in some groups it makes the bars much wider, even if I specify a width of 0.1 or similar. I find that very annoying since it makes it look very unprofessional. I want all the bars to look the same (except for the fill). I hope somebody can tell me how to fix this.

Secondly, how can I reorder the different classes in the facet windows so that the order is always C1, C2 ... C5, M, F, All where applicable. I tried it with ordering the levels of the factor, but since not all classes are present in every graph part it did not work, or at least I assume that was the reason.

Thirdly, how can I reduce the space between the bars? So that the whole graph is more compressed. Even if I make the image smaller for exporting, R will scale the bars smaller but the spaces between the bars are still huge.

I would appreciate feedback for any of those answers!

My Data: http://pastebin.com/embed_iframe.php?i=kNVnmcR1

My Code:

library(dplyr)
library(gdata)
library(ggplot2)
library(directlabels)
library(scales) 

all<-read.xls('all_auto_visual_c.xls')

all$station<-as.factor(all$station)
#all$group.new<-factor(all$group, levels=c('C. hyperboreus','C. glacialis','Special Calanus','M. longa','Pseudocalanus sp.','Copepoda'))

allp <- ggplot(data = all, aes(x=shortname2, y=perc_correct, group=group,fill=sample_size)) + 

  geom_bar(aes(fill=sample_size),stat="identity", position="dodge", width=0.1, colour="NA") + scale_fill_gradient("Sample size (n)",low="lightblue",high="navyblue")+
  facet_wrap(group~station,ncol=2,scales="free_x")+

  xlab("Species and stages") + ylab("Automatic identification and visual validation concur (%)") +
  ggtitle("Visual validation of predictions") + 
  theme_bw() + 

  theme(plot.title = element_text(lineheight=.8, face="bold", size=20,vjust=1), axis.text.x = element_text(colour="grey20",size=12,angle=0,hjust=.5,vjust=.5,face="bold"), axis.text.y = element_text(colour="grey20",size=12,angle=0,hjust=1,vjust=0,face="bold"),  axis.title.x = element_text(colour="grey20",size=15,angle=0,hjust=.5,vjust=0,face="bold"), axis.title.y = element_text(colour="grey20",size=15,angle=90,hjust=.5,vjust=1,face="bold"),legend.position="none", strip.text.x = element_text(size = 12, face="bold", colour = "black", angle = 0), strip.text.y = element_text(size = 12, face="bold", colour = "black"))
allp

#ggsave(allp, file="auto_visual_stackover.jpeg", height= 11, width= 8.5, dpi= 400,)

The current graph that needs some fixing:

Thanks a lot!

Erebus

1. will be rough with geom_bar, width is interpreted as a proportion of the distance between adjacent bars. Maybe use geom_segment instead, then size will be interpreted as absolute. For 2, reordering the factor levels is the way to do it. For 3, increase width, you've set it to 0.1 so you're telling ggplot to only use 10% of the space (or use geom_segment). – Hurter

Thanks! I will look into that. A question regarding the factor levels. I can put all classes (C1....C2&1 etc.) in one leveled factor even if not of all of them are present in the same facets? – Erebus

Answer 1

Assuming the bar widths are inversely proportional to the number of x-breaks, an appropriate scaling factor can be entered as a width aesthetic to control the width of the bars. But first, calculate the number of x-breaks in each panel, calculate the scaling factor, and put them back into the "all" data frame.

Updating to ggplot2 2.0.0 Each column mentioned in facet_wrap gets its own line in the strip. In the edit, a new label variable is setup in the dataframe so that the strip label remains on one line.

library(ggplot2)
library(plyr)

all = structure(list(station = structure(c(2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), .Label = c("Station 101", 
"Station 126"), class = "factor"), shortname2 = structure(c(2L, 
7L, 8L, 11L, 1L, 5L, 7L, 8L, 11L, 1L, 2L, 3L, 5L, 7L, 8L, 12L, 
11L, 1L, 6L, 8L, 15L, 14L, 9L, 10L, 4L, 6L, 2L, 7L, 8L, 11L, 
1L, 5L, 7L, 8L, 11L, 1L, 2L, 3L, 5L, 7L, 8L, 12L, 11L, 1L, 8L, 
11L, 1L, 15L, 14L, 13L, 9L, 10L), .Label = c("All", "C1", "C2", 
"C2&1", "C3", "C3&2", "C4", "C5", "Cegg", "Cnaup", "F", "M", 
"Micro", "Oith", "Tric"), class = "factor"), color = c(1L, 2L, 
3L, 4L, 5L, 6L, 7L, 8L, 10L, 11L, 12L, 13L, 14L, 15L, 16L, 17L, 
18L, 19L, 21L, 26L, 30L, 31L, 33L, 34L, 20L, 21L, 1L, 2L, 3L, 
4L, 5L, 6L, 7L, 8L, 10L, 11L, 12L, 13L, 14L, 15L, 16L, 17L, 18L, 
19L, 26L, 28L, 29L, 30L, 31L, 32L, 33L, 34L), group = structure(c(1L, 
1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 
4L, 6L, 5L, 3L, 3L, 3L, 3L, 6L, 6L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 
2L, 2L, 2L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 5L, 5L, 5L, 3L, 3L, 
3L, 3L, 3L), .Label = c("cgla", "Chyp", "Cope", "mlong", "pseudo", 
"specC"), class = "factor"), sample_size = c(11L, 37L, 55L, 16L, 
119L, 21L, 55L, 42L, 40L, 158L, 24L, 16L, 17L, 27L, 14L, 45L, 
98L, 241L, 30L, 34L, 51L, 22L, 14L, 47L, 13L, 41L, 24L, 41L, 
74L, 20L, 159L, 18L, 100L, 32L, 29L, 184L, 31L, 17L, 27L, 23L, 
21L, 17L, 49L, 185L, 30L, 16L, 46L, 57L, 16L, 12L, 30L, 42L), 
    perc_correct = c(91L, 78L, 89L, 81L, 85L, 90L, 91L, 93L, 
    80L, 89L, 75L, 75L, 76L, 81L, 86L, 76L, 79L, 78L, 90L, 97L, 
    75L, 86L, 93L, 74L, 85L, 88L, 88L, 90L, 92L, 90L, 91L, 89L, 
    89L, 91L, 90L, 89L, 81L, 88L, 74L, 78L, 90L, 82L, 84L, 82L, 
    90L, 94L, 91L, 81L, 69L, 83L, 90L, 81L)), class = "data.frame", row.names = c(NA, 
-52L))

all$station <- as.factor(all$station)

# Calculate scaling factor and insert into data frame
library(plyr)
N = ddply(all, .(station, group), function(x) length(row.names(x)))
N$Fac = N$V1 / max(N$V1)
all = merge(all, N[,-3], by = c("station", "group"))
all$label = paste(all$group, all$station, sep = ", ")


allp <- ggplot(data = all, aes(x=shortname2, y=perc_correct, group=group, fill=sample_size, width = .5*Fac)) + 
  geom_bar(stat="identity", position="dodge",  colour="NA") +
  scale_fill_gradient("Sample size (n)",low="lightblue",high="navyblue")+
  facet_wrap(~label,ncol=2,scales="free_x")   +
  xlab("Species and stages") + ylab("Automatic identification and visual validation concur (%)") +
  ggtitle("Visual validation of predictions") + 
  theme_bw() + 
  theme(plot.title = element_text(lineheight=.8, face="bold", size=20,vjust=1),
    axis.text.x = element_text(colour="grey20",size=12,angle=0,hjust=.5,vjust=.5,face="bold"), 
    axis.text.y = element_text(colour="grey20",size=12,angle=0,hjust=1,vjust=0,face="bold"), 
    axis.title.x = element_text(colour="grey20",size=15,angle=0,hjust=.5,vjust=0,face="bold"), 
    axis.title.y = element_text(colour="grey20",size=15,angle=90,hjust=.5,vjust=1,face="bold"),
    legend.position="none", 
    strip.text.x = element_text(size = 12, face="bold", colour = "black", angle = 0), 
    strip.text.y = element_text(size = 12, face="bold", colour = "black"))

allp

Coralyn

So I tried the same thing, but got the warning (with output still with different widths): Warning: Ignoring unknown aesthetics: width. CODE:pastebin.com/10U5ydai Can you help out? – Angele

@urwaCFC When I run the code above, using current CRAN version of ggplot2 (version 2.2.1), I get the chart as shown above (i.e., all bar widths are the same). I also get the warning message. If you are concerned about the warning, you might need to ask a new question. – Coralyn

@urwaCFC However, if I move the width aesthetic from geom_bar back into the global aesthetics, no warning message is generated. See edited code above. – Coralyn

Answer 2

Here what I did after suggestion from Gregor. Using geom_segment and geom_point makes a nice graph as I think.

library(ggplot2)

all<-read.xls('all_auto_visual_c.xls')

all$station<-as.factor(all$station)
all$group.new<-factor(all$group, levels=c('C. hyperboreus','C. glacialis','Combined','M. longa','Pseudocalanus sp.','Copepoda'))
all$shortname2.new<-factor(all$shortname2, levels=c('All','F','M','C5','C4','C3','C2','C1','Micro',     'Oith','Tric','Cegg','Cnaup','C3&2','C2&1'))

allp<-ggplot(all, aes(x=perc_correct, y=shortname2.new)) +
  geom_segment(aes(yend=shortname2.new), xend=0, colour="grey50") +
  geom_point(size=4, aes(colour=sample_size)) +
  scale_colour_gradient("Sample size (n)",low="lightblue",high="navyblue") +
  geom_text(aes(label = perc_correct, hjust = -0.5)) +
  theme_bw() +
  theme(panel.grid.major.y = element_blank()) +
  facet_grid(group.new~station,scales="free_y",space="free") +
  xlab("Automatic identification and visual validation concur (%)") + ylab("Species and stages")+
  ggtitle("Visual validation of predictions")+
  theme_bw() + 
  theme(plot.title = element_text(lineheight=.8, face="bold", size=20,vjust=1), axis.text.x = element_text(colour="grey20",size=12,angle=0,hjust=.5,vjust=.5,face="bold"), axis.text.y = element_text(colour="grey20",size=12,angle=0,hjust=1,vjust=0,face="bold"), axis.title.x = element_text(colour="grey20",size=15,angle=0,hjust=.5,vjust=0,face="bold"), axis.title.y = element_text(colour="grey20",size=15,angle=90,hjust=.5,vjust=1,face="bold"),legend.position="none", strip.text.x = element_text(size = 12, face="bold", colour = "black", angle = 0), strip.text.y = element_text(size = 8, face="bold", colour = "black"))

allp

ggsave(allp, file="auto_visual_no_label.jpeg", height= 11, width= 8.5, dpi= 400,)

This is what it produces!

Erebus

Nice work! Thanks for posting your own answer! I like the "dotplot" style as you did here, but if you prefer bars you should be able to fatten up the geom_segments by setting size very big. – Hurter

Thanks! Yeah, you could but in this case I like that it is slim and not distracting. The colour fill in the dot is sufficient I guess, although the colour scale is not super for plotting.... – Erebus