blog/graphics-in-r-with-ggplot2/

[utterances-bot]

Graphics in R with ggplot2 - Stats and R

Learn how to create professional graphics and plots in R (histogram, barplot, boxplot, scatter plot, line plot, density plot, etc.) with the ggplot2 package

https://statsandr.com/blog/graphics-in-r-with-ggplot2/

[AntoineSoetewey]

Comment written by Souradip Mookerjee on August 22, 2020 12:09:04:

Great article!

[AntoineSoetewey]

Comment written by Souradip Mookerjee on August 22, 2020 12:09:04:

Great article!

Comment written by Antoine Soetewey on August 22, 2020 12:11:35:

Thanks Souradip for your feedback! (and your help about commento)

[AntoineSoetewey]

Comment written by MiRcea Sauciuc on September 02, 2020 06:56:00:

Hi, Antoine,

Newer to your blog; love the content and your explanations. I'm curious about Axis Ticks: can that also be used for dates? I find myself in situations with continuous data (daily dates), and if I try to do any high-level EDA, the x-axis is jam-packed with (unreadable) dates. So I was looking for ways to fix the issue. Perhaps even making the data weekly/monthly.

Thank you!

[AntoineSoetewey]

Comment written by MiRcea Sauciuc on September 02, 2020 06:56:00:

Hi, Antoine,

Newer to your blog; love the content and your explanations. I'm curious about Axis Ticks: can that also be used for dates? I find myself in situations with continuous data (daily dates), and if I try to do any high-level EDA, the x-axis is jam-packed with (unreadable) dates. So I was looking for ways to fix the issue. Perhaps even making the data weekly/monthly.

Thank you!

Comment written by Antoine Soetewey on September 02, 2020 17:32:10:

Dear Mircea,

Good question. I'll update the blog with a section about dates, and I'll let you know when I added it.

Regards,
Antoine

[AntoineSoetewey]

Comment written by Antoine Soetewey on September 02, 2020 17:32:10:

Dear Mircea,

Good question. I'll update the blog with a section about dates, and I'll let you know when I added it.

Regards,
Antoine

Comment written by Antoine Soetewey on September 03, 2020 06:52:03:

Dear Mircea,

See this section on how to manage dates in ggplot2.

Hope this helps.

Regards,
Antoine

[AntoineSoetewey]

Comment written by Antoine Soetewey on September 03, 2020 06:52:03:

Dear Mircea,

See this section on how to manage dates in ggplot2.

Hope this helps.

Regards,
Antoine

Comment written by MiRcea Sauciuc on September 10, 2020 05:50:52:

Hello, Antoine,

Thank you very much!

Best,
Mircea

[AntoineSoetewey]

Comment written by Mostafa Mohammad Rezaee on November 24, 2020 04:37:27:

Great post. Thanks for sharing. I wish I had seen this post earlier.

I have "Statistical Graphics" this semester through the "Graphics for Statistics and Data Analysis with R", Second Edition, as the textbook.  Anyway, this book provides readers with all codes for its graphs through its website. A combination of this post and the book would be great.

[AntoineSoetewey]

Comment written by Mostafa Mohammad Rezaee on November 24, 2020 04:37:27:

Great post. Thanks for sharing. I wish I had seen this post earlier.

I have "Statistical Graphics" this semester through the "Graphics for Statistics and Data Analysis with R", Second Edition, as the textbook.  Anyway, this book provides readers with all codes for its graphs through its website. A combination of this post and the book would be great.

Comment written by Antoine Soetewey on November 24, 2020 07:17:50:

Dear Mostafa,

Glad you like it! Regarding the textbook, thanks for the suggestion too--I'll make sure to check it out.

Regards,
Antoine

[Iam-MFawad]

Hello Antoine Soetewey,
I hope you will be fine, My name is Muhammad Fawad and I am a beginner to R programming. I have code for a graph(L-ratio diagram), and I want to convert that code to ggplot2, using the theme, colors, and other features. I mean, I want to add the feature of the ggplot2 package to my code. I will be very thankful for this kind of act.

lmomentdiagram<-function(){
tau3=seq(-1,1,.01)
tau4_gev=0.10701+0.1109*tau3+0.84838*(tau3)^2+(-0.06669)*(tau3)^3+0.00567*(tau3)^4+(-0.04208)*(tau3)^5+0.03673*(tau3)^6+0*(tau3)^7+0*(tau3)^8
tau4_gpa=0+0.20196*tau3+0.95924*(tau3)^2+(-0.20096)*(tau3)^3+0.04061*(tau3)^4+0*(tau3)^5+0*(tau3)^6+0*(tau3)^7+0*(tau3)^8
tau4_glo=0.16667+0*tau3+0.83333*(tau3)^2
par(mar=c(5, 4, 2, 2))
plot(tau3,tau4_gev,type="l",col="royalblue",lwd=2,axes=FALSE,ann=FALSE,ylim=range(-.2,1),xlim=range(-1,1))
lines(tau3,tau4_gpa,col="tomato4",lty=1,lwd=1.5)
lines(tau3,tau4_glo,col="green4",lty=1,lwd=1.5)

box(lwd=2)
axis(1, las=1, at=seq(-1,1,.2),tck=-.01)
axis(2, las=2, at=seq(-.2,1,.2),tck=-.01)
abline(v=seq(-1,1,.2),col="black")
abline(h=seq(-.2,1,.2),col="black")
title(main=" Fig1 ",xlab="L-skewness",ylab="L-kurtosis")
legend("bottomleft", c("Distributions","GEV","GPA", "GLO"), col = c("black","royalblue","tomato4","green4"),text.col = c("blue","green4","green4","green4"),lwd=c(-1,2,1.5,1.5), lty = c(-1,1,1,1), pch = c(-1,-1,-1, -1),
merge = TRUE,cex=0.8, bg = 'gray90')
#legend("bottomright", col = c("darkmagenta"),text.col = "green4",lwd=c(-1), lty = c(-1), pch = c(11),merge = TRUE,cex=0.8, bg = 'gray90')
}

lmomentdiagram()
x<-c (0.11,0.154,0.144,0.676,0.333,0.664,0.789,0.234)
y<-c(0.222,0.366,0.335,0.551,0.590,0.233,0.674,0.666) 
points(x,y,pch=9,col="blue")

[AntoineSoetewey]

Hello Antoine Soetewey,
I hope you will be fine, My name is Muhammad Fawad and I am a beginner to R programming. I have code for a graph(L-ratio diagram), and I want to convert that code to ggplot2, using the theme, colors, and other features. I mean, I want to add the feature of the ggplot2 package to my code. I will be very thankful for this kind of act.

lmomentdiagram<-function(){
tau3=seq(-1,1,.01)
tau4_gev=0.10701+0.1109_tau3+0.84838_(tau3)^2+(-0.06669)_(tau3)^3+0.00567_(tau3)^4+(-0.04208)_(tau3)^5+0.03673_(tau3)^6+0*(tau3)^7+0*(tau3)^8
tau4_gpa=0+0.20196_tau3+0.95924_(tau3)^2+(-0.20096)_(tau3)^3+0.04061_(tau3)^4+0*(tau3)^5+0*(tau3)^6+0*(tau3)^7+0*(tau3)^8
tau4_glo=0.16667+0_tau3+0.83333_(tau3)^2
par(mar=c(5, 4, 2, 2))
plot(tau3,tau4_gev,type="l",col="royalblue",lwd=2,axes=FALSE,ann=FALSE,ylim=range(-.2,1),xlim=range(-1,1))
lines(tau3,tau4_gpa,col="tomato4",lty=1,lwd=1.5)
lines(tau3,tau4_glo,col="green4",lty=1,lwd=1.5)

box(lwd=2)
axis(1, las=1, at=seq(-1,1,.2),tck=-.01)
axis(2, las=2, at=seq(-.2,1,.2),tck=-.01)
abline(v=seq(-1,1,.2),col="black")
abline(h=seq(-.2,1,.2),col="black")
title(main=" Fig1 ",xlab="L-skewness",ylab="L-kurtosis")
legend("bottomleft", c("Distributions","GEV","GPA", "GLO"), col = c("black","royalblue","tomato4","green4"),text.col = c("blue","green4","green4","green4"),lwd=c(-1,2,1.5,1.5), lty = c(-1,1,1,1), pch = c(-1,-1,-1, -1),
merge = TRUE,cex=0.8, bg = 'gray90')
#legend("bottomright", col = c("darkmagenta"),text.col = "green4",lwd=c(-1), lty = c(-1), pch = c(11),merge = TRUE,cex=0.8, bg = 'gray90')
}

lmomentdiagram()
x<-c (0.11,0.154,0.144,0.676,0.333,0.664,0.789,0.234)
y<-c(0.222,0.366,0.335,0.551,0.590,0.233,0.674,0.666)
points(x,y,pch=9,col="blue")

Dear Muhammad,

Thanks for your comment.

Here is your plot (with base R) versus my plot (with ggplot2):

You'll find the code here. Feel free to edit the plot according to your needs. Also, it may be a good idea to check the tau vectors in your lmomentdiagram() function as I needed to edit it slightly to make it work (there were some missing multiplication symbols *).

Hope this helps.

Regards,
Antoine

[AntoineSoetewey]

Comment written by George Chacko on September 16, 2020 10:36:28:

Nice primer- thanks very much.

[AntoineSoetewey]

Comment written by George Chacko on September 16, 2020 10:36:28:

Nice primer- thanks very much.

Comment written by Antoine Soetewey on September 16, 2020 11:17:39:

Glad you like it George!

[Iam-MFawad]

Dear Muhammad,

Thanks for your comment.

Here is your plot (with base R) versus my plot (with ggplot2):

You'll find the code here. Feel free to edit the plot according to your needs. Also, it may be a good idea to check the tau vectors in your lmomentdiagram() function as I needed to edit it slightly to make it work (there were some missing multiplication symbols *).

Hope this helps.

Regards,
Antoine

Thank you very much, now it is perfect.

[CoyoteGulch]

Very valuable introduction to GGPLOT2, thanks!

There are some code snippets that I do not understand in the facets section:

1. p + facet_grid(. ~ drv)
2. facet_wrap(~year)

I do not know what the (. ~ drv) and (~year) arguments are referencing.

John Orr

[AntoineSoetewey]

Very valuable introduction to GGPLOT2, thanks!

There are some code snippets that I do not understand in the facets section:

1. p + facet_grid(. ~ drv)
2. facet_wrap(~year)

I do not know what the (. ~ drv) and (~year) arguments are referencing.

John Orr

Dear John,

facet_grid() and facet_wrap() allow you to draw a plot for each level of the variable.

In this context, with facet_grid(. ~ drv) you draw a plot for each level of the variable drv, and with facet_wrap(~year) you draw a plot for each year.

Hope that helps.

Regards,
Antoine

[CoyoteGulch]

Antoine,
Thanks very much for the quick answer. Can you expand on what the "." and "~" are telling GGPLOT2?

[AntoineSoetewey]

Antoine,
Thanks very much for the quick answer. Can you expand on what the "." and "~" are telling GGPLOT2?

You could read "~ year" as follows: create a multi-panel plot with one pane per year. So "~" is similar to "per" or "by".

In practice, if you want to create a multi-panel plot based on one variable you would use facet_wrap(). If you want to create a multi-panel plot based on more than one variables you would use facet_grid() because it allows for more control over the layout.

facet_grid() follows the pattern facet_grid(row_variable ~ column_variable). So with facet_grid(year ~ drv) you will create a multi-panel plot with each year as row and each level of drv as column. But with facet_grid(. ~ drv) it's like saying "I want to create a multi-panel plot with no variable as row and each level of drv as column".

Does it make more sense now?

[CoyoteGulch]

Antoine,
Yes, it is clearer now.

Thanks.

[Reineany]

Dear Antoine, as always your tutorials are clear worded, concise and yet so dense with information. There really should be something on the internet like votes one can give for the best internet tutors so that one could boost your work!

Antoine, I remain with but one question: How do I create boxplots to detect univariate outlier for a df that contains multiple variables? Do not look for the panel function but much rather something like the correlogram you once blogged about.
Isn't it possible to write a code wherein one defines the df, then says "compute a boxplot for all variables / multiple variables", colour outliers and put alll those in an overview (like with the pairs plot).
Also I faintly recall that you had suggestions on how to have boxplots on the left and qqplots on the right? I would be really grateful if you could point out some advice on this!
your fan from Germany

[AntoineSoetewey]

Dear @Reineany,

First, thanks for your kind feedback!

How do I create boxplots to detect univariate outlier for a df that contains multiple variables? Do not look for the panel function but much rather something like the correlogram you once blogged about.

The correlogram (see the post here) will not tell you whether there are outliers or not, so I suppose you want to create a boxplot for each variable of the dataset. If you have a limited number of variables, you can simply draw a boxplot manually for each variable (with boxplot(dat$var) or its ggplot2 counterpart). If you have many variables, you can create a boxplot in a for loop, as follows:

# packages
library(ggplot2)

# create simulated data
dat <- data.frame(
  var1 = rnorm(1000),
  var2 = rnorm(1000),
  var3 = rnorm(1000)
)

# plots
for (i in 1:ncol(dat)) {
  print(ggplot(dat, aes(y = dat[, i])) +
    geom_boxplot(outlier.color = "steelblue") +
    labs(y = names(dat)[i]) +
    theme_minimal())
}

Isn't it possible to write a code wherein one defines the df, then says "compute a boxplot for all variables / multiple variables", colour outliers and put alll those in an overview (like with the pairs plot).

The above code allows to draw a boxplot for all variables of the dataset at once. Colouring outliers is done with outlier.color = "steelblue". However, I don't see what you mean by putting all outliers in an overview. Do you have any example so I can have a look?

Also I faintly recall that you had suggestions on how to have boxplots on the left and qqplots on the right? I would be really grateful if you could point out some advice on this!

I don't remember where (and if) I suggested that, but anyway, here is how:

# packages
library(ggplot2)
library(patchwork)
library(qqplotr)

# create simulated data
dat <- data.frame(
  var = rnorm(1000)
)

# save plots
p1 <- ggplot(dat, aes(y = var)) +
  geom_boxplot() +
  theme_minimal()
p2 <- ggplot(dat, aes(sample = var)) +
  stat_qq_band() +
  stat_qq_line() +
  stat_qq_point() +
  theme_minimal()

# put plots next to each other
p1 + p2

which gives:

(Note that if you want to visualize the normality, I recommend plotting a histogram and a QQ-plot, instead of a boxplot and a QQ-plot.)

Hope this helps.

Regards,
Antoine

[berry-blue01]

Hello. Thank you for the great article! I had a question regarding the line charts. I would like to make one to track performance, but the problem is that my the variable on the x-achse is a factor. Is it possible to bild a line (trend) chart, when one variable is numeric and the other one a factor?

• Thank you in advance!

Keep up the good work :)

[AntoineSoetewey]

Dear @berry-blue01,

If you have only one observation per group, yes you can draw a line plot chart with x as factor. For example:

dat <- data.frame(x = c("A", "B", "C"),
                  y = c(1.4, 2.7, 4.9))

library(ggplot2)

ggplot(dat) +
  aes(x = x, y = y, group = 1) +
  geom_line()

However, if you have several observations per factor, I recommend:

• computing the mean or median by factor and plot it as a line chart
• build a boxplot for each factor (see the second plot in this section)

Hope this helps.

Regards,
Antoine

[berry-blue01]

Thank you so much! It's a huge help. I really appreciate you answering :)