README.md

In IngallsLabUW/ggbarribbon: Interpolated Stacked Bar Plots with ggplot2

ggbarribbon

A small repo that implements a new geom for ggplot2: geom_bar_ribbon. This geom interpolates smoothly between the values that would normally be provided to geom_bar using a sigmoid curve.

Often, this is useful when a stacked bar plot is desired but the values along the perpindicular axis are numeric rather than discrete. A numeric axis often results in overplotting or weird spacing issues when the data could instead be interpolated and show up more nicely.

# devtools::install_github("https://github.com/IngallsLabUW/ggbarribbon")
library(ggbarribbon)

metab_df <- data.frame(
  metab=LETTERS[1:10], 
  depth=rep(c(0, 5, 25, 125, 175), each=10),
  area=runif(50)
)

For example, the barplot below does a nice job of conveying changes in relative metabolite abundance over depth, but the spacing on the y axis is awful and results in a large amount of whitespace.

library(ggplot2)
ggbar <- ggplot(metab_df) +
  geom_bar(aes(y=area, x=-depth, fill=metab), stat = "identity", 
           position = position_fill(reverse = TRUE), color="black") +
  coord_flip()
ggbar

Instead, with geom_bar_ribbon, the data can be smoothly interpolated so that the whitespace is used and the trends between different depths are much clearer.

ggbarribbon <- ggplot(metab_df) +
  geom_bar_ribbon(aes(y=area, x=-depth, fill=metab), color="black") +
  coord_flip()
ggbarribbon

A direct comparison:

gridExtra::grid.arrange(ggbar, ggbarribbon)

And of course, because it’s built within the ggplot2 family it’s completely customizable:

ggbarribbon +
  geom_vline(xintercept = -metab_df$depth, lwd=1) +
  scale_y_continuous(labels = scales::percent) +
  theme_minimal() +
  theme(legend.position = "top",
        text = element_text(size = 18)) +
  labs(x="Depth (m)", y="Metabolite proportion", fill="Metabolite ID") +
  scale_fill_manual(values = LaCroixColoR::lacroix_palette("Pamplemousse", type="paired", n = 10))

As another example, below I’ve recreated the plots of Figure 4 from Heal et al., (2020)

options(readr.read_lazy = FALSE, readr.show_col_types = FALSE)
# Override normal read.csv to access correct standards list
old_stan_csv_url <- "https://raw.githubusercontent.com/IngallsLabUW/Ingalls_Standards/3e463eaa2cfd6d58a8b338a99b4d965f9a4d492e/Ingalls_Lab_Standards_NEW.csv"
read.csv <- function(text, header){
  utils::read.csv(old_stan_csv_url, header = TRUE)
}

# Download necessary files and mimic project workspace
base_url <- "https://raw.githubusercontent.com/kheal/Gradients1_SemiTargeted3"
sha <- "234f1753867385a978063f96cc6d9c0c67c248af"

dir.create("Intermediates")
longdat_file <- "Intermediates/Quantified_LongDat_Enviro.csv"
download.file(paste(base_url, sha, longdat_file, sep = "/"), destfile = longdat_file)
dir.create("MetaData")
metadat_file <- "MetaData/SampInfo_wMetaData_withUTC.csv"
download.file(paste(base_url, sha, metadat_file, sep = "/"), destfile = metadat_file)

dir.create("Figures/Manuscript_figures", recursive = TRUE)

source("https://raw.githubusercontent.com/kheal/Gradients1_SemiTargeted3/master/Figures/Code/BarPlots_molPerCent_C.R")

b_c_d.big

unlink("Intermediates", recursive = TRUE)
unlink("MetaData", recursive = TRUE)
unlink("Figures", recursive = TRUE)

Instead, we can now use geom_bar_ribbon to interpolate smoothly between the discrete data points. All that I’ve done with the code below is to copy the original code directly and replace calls to geom_bar with calls to geom_bar_ribbon. I then coerced the x-values out of factor form because they can now be plotted as continuous variables. I then also added the new interp_res argument to the first call because its default is a single unit value (works great for depth profiles, not for latitude).

AllSmps_long_Summary_Tran <- AllSmps_long_Summary_withOthers %>%
  filter(Cruise == "KOK1606") %>% 
  # Line below is new
  mutate(Latitude=as.numeric(levels(Latitude))[Latitude])

b.big <- ggplot(AllSmps_long_Summary_Tran, aes(x = Latitude, y = molFractionC, fill = MF_Frac))+
  # Line below is new
  geom_bar_ribbon(stat = "identity", color = "black", size = 0.2, interp_res = 10)+
  scale_y_continuous(expand = c(0, 0), limits = c(0,1))+
  scale_fill_manual(values = as.character(unique(AllSmps_long_Summary_Tran$Colors)),
                    guide = guide_legend(nrow=5, keyheight = 1, keywidth = 1 ))+
  labs(y= "mole fraction C")+
  theme(legend.title = element_blank(),
        legend.text = element_text(size = 6),
        legend.position="bottom",
        axis.title = element_text(size = 9),
        axis.text = element_text(size = 8))+
  coord_flip() 

#Plotting up the North Depth Profile -----
AllSmps_long_Summary_nDP <- AllSmps_long_Summary_withOthers %>%
  filter(Cruise ==  "MGL1704") %>%
  unique()

c.big <- ggplot(AllSmps_long_Summary_nDP, aes(x = as.numeric(Depth), y = molFractionC, fill = MF_Frac))+
  scale_x_reverse(breaks = rev(unique(as.numeric(AllSmps_long_Summary_nDP$Depth)))) +
  scale_y_continuous(expand = c(0, 0), 
                     limits = c(0,1))+
  geom_bar_ribbon(stat = "identity", color = "black", width = 15, size = 0.2) +
  scale_fill_manual(values = as.character(unique(AllSmps_long_Summary_nDP$Colors)),
                    guide = guide_legend(nrow=5,keyheight = 0.8, keywidth = 0.8 ))+
  labs(y= "mole fraction C", x = "Depth (m)")+
  guides(fill=FALSE) +
  theme(axis.title = element_text(size = 9),
        axis.text = element_text(size = 8))+
  coord_flip() 

#Plotting up the South Depth Profile -----
AllSmps_long_Summary_sDP <- AllSmps_long_Summary_withOthers %>%
  filter(Cruise ==  "KM1513") %>%
  unique()

d.big <- ggplot(AllSmps_long_Summary_sDP, aes(x = as.numeric(Depth), y = molFractionC, fill = MF_Frac))+
  scale_x_reverse(breaks = rev(unique(as.numeric(AllSmps_long_Summary_sDP$Depth)))) +
  scale_y_continuous(expand = c(0, 0), limits = c(0,1))+
  geom_bar_ribbon(stat = "identity", color = "black", width = 15, size = 0.2) +
  scale_fill_manual(values = as.character(unique(AllSmps_long_Summary_Tran$Colors)),
                    guide = guide_legend(nrow=5,keyheight = 0.8, keywidth = 0.8 ))+
  labs(y= "mole fraction C", x = "Depth (m)")+
  guides(fill=FALSE) +
  theme(axis.title = element_text(size = 9),
        axis.text = element_text(size = 8))+
  coord_flip() 

#Combine the plots 
c_d2.big <- plot_grid(c.big, d.big, ncol = 1, labels = c('', ''), label_fontface = "bold")
b_c_d.big <- plot_grid(b.big, c_d2.big, ncol = 2, labels = c('', ''), label_fontface = "bold")
b_c_d.big

Interestingly, a few values are now shown to have hidden NAs (resulting in the white patches), but the plots as a whole behave similarly. One other note is that the plots arrange their fills from largest to smallest, left to right, rather than the ggplot2 default of right to left.

README last compiled on 2021-11-15.

IngallsLabUW/ggbarribbon documentation built on Dec. 17, 2021, 11:29 p.m.