Having Titanic dataset I draw histograms of age/sex against the passenger class.
str(titanic) gives the following data
> 'data.frame': 714 obs. of 4 variables:
$ Survived: int 0 1 1 1 0 0 0 1 1 1 ...
$ Pclass : int 3 1 3 1 3 1 3 3 2 3 ...
$ Sex : chr "male" "female" "female" "female" ...
$ Age : num 22 38 26 35 35 54 2 27 14 4 ...
First, I made a plot of proportion of male/female against the travel classes.
It has been done by
ggplot(data = titanic, aes(x = factor(Age), fill = factor(Sex))) +
geom_bar(position = "dodge", aes(y = (..count..)/sum(..count..))) +
facet_grid(. ~ Pclass) + scale_x_discrete(breaks=c(20,40,60)) +
ylab("Frequency") + xlab("Age") +
scale_fill_discrete(name = "Sex")
Now I want to use the same graph, but add additional information -> proportion of survivals for all categories.
For example, what is the proportion of survival women age 20-30 who traveled the first class.
I would like to see it in the same bars, i.e split each column into two parts(survived/not survived).
Can I do it with ggplot? And if yes, how?
Using the builtin Titanic data set, I can show you roughly what #Axeman suggested in the comments. Note that it only has two categories for age (Child/Adult) so you would need to decide how to bin for your data.
ggplot(as.data.frame(Titanic)
, aes(y = Freq
, x = Age
, fill = Survived)) +
geom_col() +
facet_grid(Sex ~ Class)
Importantly, I am not sure that you are gaining anything by showing the frequencies in the way you currently are, as they do not appear to be showing anything meaningfully different than the counts would. If, instead, you wanted to show the proportion within each group that survived, you may be better off calculating those percentages first, then passing them to ggplot. Here is an example of that using dplyr. Again, your age bins can be whatever you want, but note that the narrower the bins, the more noisy the data will be.
as.data.frame(Titanic) %>%
group_by(Class, Sex, Age) %>%
mutate(Proportion = Freq/ sum(Freq)) %>%
ggplot(aes(y = Proportion
, x = Age
, fill = Survived)) +
geom_col() +
facet_grid(Sex ~ Class)
Related
There seems to be a problem with R's ggplot2 library when I include both the fill and group parameters in a bar plot (geom_bar()). I've already tried looking for answers for several hours but couldn't find one that would help. This is actually my first post here.
To give a little background, I have a dataframe named smokement (short for smoke and mental health), a categorical variable named smoke100 (smoked in the past 100 days?) with "Yes" and "No", and another categorical variable named misnervs (frequency of feelings of nervousness) with 5 possible values: "All", "Most", "Some", "A little", and "None."
When I run this code, I get this result:
ggplot(data = smokement) +
geom_bar(aes(x = smoke100, fill = smoke100)) +
facet_wrap(~misnervs, nrow = 1)
However, the result I want is to have all grouped bar plots display their respective proportions. By reading a bit of "R for Data Science" book I found out that I need to include y = ..prop.. and group = 1 in aes() to achieve it:
ggplot(data = smokement) +
geom_bar(aes(x = smoke100, y = ..prop.., group = 1)) +
facet_wrap(~misnervs, nrow = 1)
Finally, I try to use the fill = smoke100 parameter in aes() to display this categorical variable in color, just like I did on the first code. But when I add this fill parameter, it doesn't work! The code runs, but it shows exactly the same output as the second code, as if the fill parameter this time was somehow ignored!
ggplot(data = smokement) +
geom_bar(aes(x = smoke100, y = ..prop.., group = 1, fill = smoke100)) +
facet_wrap(~misnervs, nrow = 1)
Does anyone have an idea of why this happens, and how to solve it? My end goal is to display each value of smoke100 (the "Yes" and "No" bars) with colors and a legend at the right, just like on the first graph, while having each grouping level of "misnervs" display their respective proportions of smoke100 ("Yes", "No") levels, just like on the second graph.
EDIT:
> dim(smokement)
[1] 35471 6
> str(smokement)
'data.frame': 35471 obs. of 6 variables:
$ smoke100: Factor w/ 2 levels "Yes","No": 1 2 1 2 1 1 1 1 1 1 ...
$ misnervs: Factor w/ 5 levels "All","Most","Some",..: 3 4 5 4 1 5 3 3 5 5 ...
$ mishopls: Factor w/ 5 levels "All","Most","Some",..: 3 5 5 5 5 5 5 5 5 5 ...
$ misrstls: Factor w/ 5 levels "All","Most","Some",..: 3 5 5 3 1 5 3 5 1 5 ...
$ misdeprd: Factor w/ 5 levels "All","Most","Some",..: 5 5 5 5 4 5 5 5 5 5 ...
$ miswtles: Factor w/ 5 levels "All","Most","Some",..: 5 5 5 5 5 5 5 5 5 5 ...
> head(smokement)
smoke100 misnervs mishopls misrstls misdeprd miswtles
1 Yes Some Some Some None None
2 No A little None None None None
3 Yes None None None None None
4 No A little None Some None None
5 Yes All None All A little None
6 Yes None None None None None
As for the output without group = 1
ggplot(data = smokement) +
+ geom_bar(aes(x = smoke100, y = ..prop.., fill = smoke100)) +
+ facet_wrap(~misnervs, nrow = 1)
Besides the solution offered here the GGAlly package includes a stat_prop which introduces a new by aesthetic to specify the way the proportions should be calculated:
library(GGally)
ggplot(data = smokement) +
geom_bar(aes(x = smoke100, y = ..prop.., fill = smoke100, by = misnervs), stat = "prop") +
facet_wrap(~misnervs, nrow = 1)
And just for reference the same could be achieved without GGAlly by setting fill=factor(..x..):
ggplot(data = smokement) +
geom_bar(aes(x = smoke100, y = ..prop.., fill = factor(..x..), group = 1)) +
facet_wrap(~misnervs, nrow = 1)
DATA
misnervs <- c("All", "Most", "Some", "A little", "None")
set.seed(123)
smokement <-
data.frame(
smoke100 = sample(c("Yes", "No"), 100, replace = TRUE),
misnervs = factor(sample(misnervs, 100, replace = TRUE), levels = misnervs)
)
I wasn't able to get what you wanted by tweaking your call to geom_bar*, but I think this gives you what you are looking for. As you didn't provide your input dataset (for understandable reasons), I've used the diamonds tibble in my code. The changes you need to make should be obvious.
*: I'm sure it can be done: I just wasn't able to work it out.
The idea behind my solution is to pre-compute the proportions you want to plot before the call to ggplot.
group_modify takes a grouped tibble and applies the specified function to each group in turn, before returning the modified (grouped) tibble.
diamonds %>%
group_by(cut) %>%
group_modify(
function(.x, .y)
.x %>%
group_by(color) %>%
summarise(Prop=n()/nrow(.))
) %>%
ggplot() +
geom_col(aes(x=color, y=Prop, fill=color)) +
facet_wrap(~cut)
Note the switch from geom_bar to geom_col: geom_bar uses row counts, geom_col uses values in the data.
As a rough-and-ready QC, here's the equivalent of your code that produces the "all grey' plot:
diamonds %>%
ggplot() +
geom_bar(aes(x=color, y=..prop.., fill=color, group=1)) +
facet_wrap(~cut)
I am using facet grid to generate neat presentations of my data.
Basically, my data frame has four columns:
idx, density, marker, case.
There are 5 cases, each case corresponds to 5 markers, and each marker corresponds to multiple idx, each idx corresponds to one density.
The data is uploaded here:
data frame link
I tried to use facet_grid to achieve my goal, however, I obtained a really messed up graph:
The x-axis and y-axis are messed up, the codes are:
library(ggplot2)
library(cowplot)
plot.density <-
ggplot(df_densityWindow, aes(x = idx, y = density)) +
geom_col() +
facet_grid(marker ~ case, scales = 'free') +
background_grid(major = 'y', minor = "none") + # add thin horizontal lines
panel_border() # and a border around each panel
plot(plot.density)
EDIT:
I reupload the file, now it should be work:
download file here
All 4 columns have been read as factors. This is an issue from however you loaded the data into R. Take a look at:
df <- readRDS('df.rds')
str(df)
'data.frame': 52565 obs. of 4 variables:
$ idx : Factor w/ 4712 levels "1","10","100",..: 1 1112 2223 3334 3546 3657 3768 3879 3990 2 ...
$ density: Factor w/ 250 levels "1022.22222222222",..: 205 205 204 203 202 201 199 198 197 197 ...
$ marker : Factor w/ 5 levels "CD3","CD4","CD8",..: 1 1 1 1 1 1 1 1 1 1 ...
$ case : Factor w/ 5 levels "Case_1","Case_2",..: 1 1 1 1 1 1 1 1 1 1 ...
Good news is that you can fix it with:
df$idx <- as.integer(as.character(df$idx))
df$density <- as.numeric(as.character(df$density))
Although you should look into how you are loading the data, to avoid future.
As another trick, try the above code without using the as.character calls, and compare the differences.
As already explained by MrGumble, the idx and density variables are of type factor but should be plotted as numeric.
The type.convert() function does the data conversion in one go:
library(ggplot2)
library(cowplot)
ggplot(type.convert(df_densityWindow), aes(x = idx, y = density)) +
geom_col() +
facet_grid(marker ~ case, scales = 'free') +
background_grid(major = 'y', minor = "none") + # add thin horizontal lines
panel_border() # and a border around each panel
This question already has answers here:
Add legend to ggplot2 line plot
(4 answers)
Closed 2 years ago.
I was attempting (unsuccessfully) to show a legend in my R ggplot2 graph which involves multiple plots. My data frame df and code is as follows:
Individuals Mod.2 Mod.1 Mod.3
1 2 -0.013473145 0.010859793 -0.08914021
2 3 -0.011109863 0.009503278 -0.09049672
3 4 -0.006465788 0.011304668 -0.08869533
4 5 0.010536718 0.009110458 -0.09088954
5 6 0.015501212 0.005929766 -0.09407023
6 7 0.014565584 0.005530390 -0.09446961
7 8 -0.009712516 0.012234843 -0.08776516
8 9 -0.011282278 0.006569570 -0.09343043
9 10 -0.011330579 0.003505439 -0.09649456
str(df)
'data.frame': 9 obs. of 4 variables:
$ Individuals : num 2 3 4 5 6 7 8 9 10
$ Mod.2 : num -0.01347 -0.01111 -0.00647 0.01054 0.0155 ...
$ Mod.1 : num 0.01086 0.0095 0.0113 0.00911 0.00593 ...
$ Mod.3 : num -0.0891 -0.0905 -0.0887 -0.0909 -0.0941 ...
ggplot(df, aes(df$Individuals)) +
geom_point(aes(y=df[,2]), colour="red") + geom_line(aes(y=df[,2]), colour="red") +
geom_point(aes(y=df[,3]), colour="lightgreen") + geom_line(aes(y=df[,3]), colour="lightgreen") +
geom_point(aes(y=df[,4]), colour="darkgreen") + geom_line(aes(y=df[,4]), colour="darkgreen") +
labs(title = "Modules", x = "Number of individuals", y = "Mode")
I looked up the following stackflow threads, as well as Google searches:
Merging ggplot2 legend
ggplot2 legend not showing
`ggplot2` legend not showing label for added series
ggplot2 legend for geom_area/geom_ribbon not showing
ggplot and R: Two variables over time
ggplot legend not showing up in lift chart
Why ggplot2 legend not show in the graph
ggplot legend not showing up in lift chart.
This one was created 4 days ago
This made me realize that making legends appear is a recurring issue, despite the fact that legends usually appear automatically.
My first question is what are the causes of a legend to not appear when using ggplot? The second is how to solve these causes. One of the causes appears to be related to multiple plots and the use of aes(), but I suspect there are other reasons.
colour= XYZ should be inside the aes(),not outside:
geom_point(aes(data, colour=XYZ)) #------>legend
geom_point(aes(data),colour=XYZ) #------>no legend
Hope it helps, it took me a hell long way to figure out.
You are going about the setting of colour in completely the wrong way. You have set colour to a constant character value in multiple layers, rather than mapping it to the value of a variable in a single layer.
This is largely because your data is not "tidy" (see the following)
head(df)
x a b c
1 1 -0.71149883 2.0886033 0.3468103
2 2 -0.71122304 -2.0777620 -1.0694651
3 3 -0.27155800 0.7772972 0.6080115
4 4 -0.82038851 -1.9212633 -0.8742432
5 5 -0.71397683 1.5796136 -0.1019847
6 6 -0.02283531 -1.2957267 -0.7817367
Instead, you should reshape your data first:
df <- data.frame(x=1:10, a=rnorm(10), b=rnorm(10), c=rnorm(10))
mdf <- reshape2::melt(df, id.var = "x")
This produces a more suitable format:
head(mdf)
x variable value
1 1 a -0.71149883
2 2 a -0.71122304
3 3 a -0.27155800
4 4 a -0.82038851
5 5 a -0.71397683
6 6 a -0.02283531
This will make it much easier to use with ggplot2 in the intended way, where colour is mapped to the value of a variable:
ggplot(mdf, aes(x = x, y = value, colour = variable)) +
geom_point() +
geom_line()
ind = 1:10
my.df <- data.frame(ind, sample(-5:5,10,replace = T) ,
sample(-5:5,10,replace = T) , sample(-5:5,10,replace = T))
df <- data.frame(rep(ind,3) ,c(my.df[,2],my.df[,3],my.df[,4]),
c(rep("mod.1",10),rep("mod.2",10),rep("mod.3",10)))
colnames(df) <- c("ind","value","mod")
Your data frame should look something likes this
ind value mod
1 5 mod.1
2 -5 mod.1
3 3 mod.1
4 2 mod.1
5 -2 mod.1
6 5 mod.1
Then all you have to do is :
ggplot(df, aes(x = ind, y = value, shape = mod, color = mod)) +
geom_line() + geom_point()
I had a similar problem with the tittle, nevertheless, I found a way to show the title: you can add a layer using
ggtitle ("Name of the title that you want to show")
example:
ggplot(data=mtcars,
mapping = aes(x=hp,
fill = factor(vs)))+
geom_histogram(bins = 9,
position = 'identity',
alpha = 0.8, show.legend = T)+
labs(title = 'Horse power',
fill = 'Vs Motor',
x = 'HP',
y = 'conteo',
subtitle = 'A',
caption = 'B')+
ggtitle("Horse power")
I'm trying to make a horizontal stacked barplot using ggplot. Below are the actual values for three out of 300 sites in my data frame. Here's where I've gotten to so far, using info pulled from these previous questions which I admit I may not have fully understood.
df <- data.frame(id=c("AR001","AR001","AR001","AR001","AR002","AR002","AR002","AR003","AR003","AR003","AR003","AR003"),
landuse=c("agriculture","developed","forest","water","agriculture","developed","forest","agriculture","developed","forest","water","wetlands"),
percent=c(38.77,1.76,59.43,0.03,69.95,0.42,29.63,65.4,3.73,15.92,1.35,13.61))
df
id landuse percent
1 AR001 agriculture 38.77
2 AR001 developed 1.76
3 AR001 forest 59.43
4 AR001 water 0.03
5 AR002 agriculture 69.95
6 AR002 developed 0.42
7 AR002 forest 29.63
8 AR003 agriculture 65.40
9 AR003 developed 3.73
10 AR003 forest 15.92
11 AR003 water 1.35
12 AR003 wetlands 13.61
str(df)
'data.frame': 12 obs. of 3 variables:
$ id : Factor w/ 3 levels "AR001","AR002",..: 1 1 1 1 2 2 2 3 3 3 ...
$ landuse: Factor w/ 5 levels "agriculture",..: 1 2 3 4 1 2 3 1 2 3 ...
$ percent: num 38.77 1.76 59.43 0.03 69.95 ...
df <- transform(df,
landuse.ord = factor(
landuse,
levels=c("agriculture","forest","wetlands","water","developed"),
ordered =TRUE))
cols <- c(agriculture="maroon",forest="forestgreen",
wetlands="gold", water="dodgerblue", developed="darkorchid")
ggplot(df,aes(x = id, y = percent, fill = landuse.ord, order=landuse.ord)) +
geom_bar(position = "stack",stat = "identity", width=1) +
coord_flip() +
scale_fill_manual(values = cols)
which produces this graph.
What I would like to do is to reorder the bars so that they are in descending order by value for the agriculture category - in this example AR002 would be at the top, followed by AR003 then AR001. I tried changing the contents of aes to aes(x = reorder(landuse.ord, percent), but that eliminated the stacking and seemed to have maybe summed the percentages for each land use category:
I would like to have the stacks in order, from left to right: agriculture, forest, wetlands, water, developed. I tried doing that with the transform part of the code, which put it in the correct order in the legend, but not in the plot itself?
Thanks in advance... I have made a ton of progress based on answers to other peoples' questions, but seem to now be stuck at this point!
Update: here is the finished graph for all 326 sites!
Ok based on your comments, I believe this is your solution. Place these lines after cols<-...:
#create df to sort by argiculture's percentage
ag<-filter(df, landuse=="agriculture")
#use the df to sort and order df$id's levels
df$id<-factor(df$id, levels=ag$id[order(ag$percent)], ordered = TRUE)
#sort df, based on ordered ids and ordered landuse
df<-df[order(df$id, df$landuse.ord),]
ggplot(df,aes(x = id, y = percent, fill = landuse.ord, order=landuse.ord)) +
geom_bar(position = "stack",stat = "identity", width=1) +
coord_flip() +
scale_fill_manual(values = cols)
The comments should clarify each of the lines purposes. This will reorder your original data frame, if that is a problem I would create a copy and then operate on the new copy.
Hi Stack Overflow community,
I have a dataset:
conc branch length stage factor
1 1000 3 573.5 e14 NRG4
2 1000 7 425.5 e14 NRG4
3608 1000 44 5032.0 P10 NRG4
3609 1000 0 0.0 P10 NRG4
FYI
> str(dframe1)
'data.frame': 3940 obs. of 5 variables:
$ conc : Factor w/ 6 levels "0","1","10","100",..: 6 6 6 6 6 6 6 6 6 6 ...
$ branch: int 3 7 5 0 1 0 0 4 1 1 ...
$ length: num 574 426 204 0 481 ...
$ stage : Factor w/ 8 levels "e14","e16","e18",..: 1 1 1 1 1 1 1 1 1 1 ...
$ factor: Factor w/ 2 levels "","NRG4": 2 2 2 2 2 2 2 2 2 2 ...
I would like to create facetted line graphs, plotting the mean +/- standard error of the mean
I have tried experimenting and building a ggplot from others (here and on the web).
I have successfully used scripts that will make bargraphs this way:
errbar.ggplot.facets <- ggplot(dframe1, aes(x = conc, y = length))
### function to calculate the standard error of the mean
se <- function(x) sd(x)/sqrt(length(x))
### function to be applied to each panel/facet
my.fun <- function(x) {
data.frame(ymin = mean(x) - se(x),
ymax = mean(x) + se(x),
y = mean(x))}
g.err.f <- errbar.ggplot.facets +
stat_summary(fun.y = mean, geom = "bar",
fill = clrs.hcl(48)) +
stat_summary(fun.data = my.fun, geom = "linerange") +
facet_wrap(~ stage) +
theme_bw()
print(g.err.f)
Source: http://teachpress.environmentalinformatics-marburg.de/2013/07/creating-publication-quality-graphs-in-r-7/
In fact, I have created facetted line graphs with this script:
`ggplot(data=dframe1, aes(x=conc, y = length, group = stage)) +
geom_line() + facet_wrap(~stage)`
image: postimg.org/image/ebpdc0sb7
However, I used a transformed dataset of only means, SEM in another column, but I don't know how to add them.
Given the complexity (for me) of the bargraphs + error line scripts above, I have not yet been able to integrate/synthesize these into something I need.
In this case, the colour is not important to have.
P.S. I apologise for the long thread (and perhaps the overkill on some details). This is my first online R question, so not sure of correct etiquette. Thank you all in advance for being so helpful!
Darian
In case your dataframe has a column for the mean and the se you could do something like this:
library("dplyr")
library("ggplot2")
# Create a dummydataframe with columns mean and se
df <- mtcars %>%
group_by(gear, cyl) %>%
summarise(mean_mpg = mean(mpg), se_mpg = se(mpg))
ggplot(df, aes(x = gear, y = mean_mpg)) +
geom_bar(stat = "identity") +
geom_errorbar(aes(ymin = mean_mpg - se_mpg, ymax = mean_mpg + se_mpg)) +
facet_wrap(~cyl)