r 中带有 4 个变量的分组条形图

Question

I'm a beginner in r and I've been trying to find how I can plot this graphic.

I have 4 variables (% of gravel, % of sand, % of silt in five places). I'm trying to plot the percentages of these 3 types of sediment (y) in each station (x). So it's five groups in x axis and 3 bars per group.

   Station   % gravel    % sand      % silt
1   PRA1    28.430000   70.06000    1.507000
2   PRA3    19.515000   78.07667    2.406000
3   PRA4    19.771000   78.63333    1.598333
4   PRB1    7.010667    91.38333    1.607333
5   PRB2    18.613333   79.62000    1.762000 

I tried plotting a grouped barchart with

grao <- read_excel("~/Desktop/Masters/Data/grao.xlsx")
colors <- c('#999999','#E69F00','#56B4E9','#94A813','#718200')
barplot(table(grao$Station, grao$`% gravel`, grao$`% sand`, grao$`% silt`), beside = TRUE, col = colors)

But this error message keeps happening:

'height' must be a vector or matrix

I also tried

ggplot(grao, aes(Station, color=as.factor(`% gravel`), shape=as.factor(`% sand`))) + 
geom_bar() + scale_color_manual(values=c('#999999','#E69F00','#56B4E9','#94A813','#718200')+ theme(legend.position="top")

But it's creating a crazy graphic.

Could someone help me, please? I've been stuck for weeks now in this one.

Cheers

Answer 1

I think this may be what you are looking for:

#install.packages("tidyverse")
library(tidyverse)
df <-  data.frame(
  station = c("PRA1", "PRA3", "PRA4", "PRB1", "PRB2"),
  gravel = c(28.4, 19.5, 19.7, 7.01, 18.6),
  sand = c(70.06, 78.07, 78.63, 91, 79),
  silt = c(1.5, 2.4, 1.6, 1.7, 1.66)
)

df2 <- df %>% 
  pivot_longer(cols = c("gravel", "sand", "silt"), names_to = "Sediment_Type", values_to = "Percentage")

ggplot(df2) +
  geom_bar(aes(x = station, y = Percentage, fill = Sediment_Type ), stat = "identity", position = "dodge") +
theme_minimal() #theme_minimal() is from the ggthemes package

provides:

You need to "pivot" your data set "longer". Part of the tidy way is ensuring all columns represent a single variable. You will notice in your initial dataframe that each column name is a variable ("Sediment_type") and each column fill is just the percentage for each ("Percentage"). The function pivot_longer() takes a dataset and allows one to gather up all the columns then turn them into just two - the identity and value.

Once you've done this, ggplot will allow you to specify your x axis, and then a grouping variable by "fill". You can switch these two up. If you end up with lots of data and grouping variables, faceting is also an option worth looking in to!

Hope this helps,

Brennan

Answer 2

barplot wants a "matrix" , ideally with both dimension names. You could transform your data like this (remove first column while using it for row names):

dat <- `rownames<-`(as.matrix(grao[,-1]), grao[,1])

You will see, that barplot already does the tabulation for you. However, you also could use xtabs ( table might not be the right function for your approach).

# dat <- xtabs(cbind(X..gravel, X..sand, X..silt) ~ Station, grao)  ## alternatively

I would advise you to use proper variable names, since special characters are not the best idea.

colnames(dat) <- c("gravel", "sand", "silt")
dat
#         gravel     sand     silt
# PRA1 28.430000 70.06000 1.507000
# PRA3 19.515000 78.07667 2.406000
# PRA4 19.771000 78.63333 1.598333
# PRB1  7.010667 91.38333 1.607333
# PRB2 18.613333 79.62000 1.762000

Then barplot knows what's going on.

.col <- c('#E69F00','#56B4E9','#94A813')  ## pre-define colors
barplot(t(dat), beside=T, col=.col, ylim=c(0, 100),  ## barplot
        main="Here could be your title", xlab="sample", ylab="perc.")
legend("topleft", colnames(dat), pch=15, col=.col, cex=.9, horiz=T, bty="n")  ## legend
box()  ## put it in a box

---

Data:

grao <- read.table(text="   Station   '% gravel'    '% sand'      '% silt'
1   PRA1    28.430000   70.06000    1.507000
2   PRA3    19.515000   78.07667    2.406000
3   PRA4    19.771000   78.63333    1.598333
4   PRB1    7.010667    91.38333    1.607333
5   PRB2    18.613333   79.62000    1.762000 ", header=TRUE)