# R pairs & ggpairs Plot Functions | 5 Example Codes (Color, Labels, Panels & by Group)

Basic R Syntax:

`pairs(data)`

The pairs R function returns a plot matrix, consisting of scatterplots for each variable-combination of a data frame. The basic R syntax for the pairs command is shown above.

In the following tutorial, I’ll explain in five examples how to use the pairs function in R.

## Example 1: Basic Application of pairs() in R

I’m going to start with a very basic application of the pairs R function. Let’s first create some random data for this example:

```set.seed(525354) # Set seed for reproducibility N <- 1000 # Sample size of 1000   x1 <- rnorm(N) # Create variable x2 <- x1 + rnorm(N, 0, 3) # Create correlated variable x3 <- 2 * x1 - x2 + rnorm(N, 0, 2) # Create another correlated variable   data <- data.frame(x1, x2, x3) # Combine all variables to data.frame```

Our example data contains three numeric variables and 1,000 rows.

Now, let’s apply the pairs function in R:

`pairs(data) # Apply pairs function` Figure 1: Basic pairs() R Plot.

As you can see, we are able to produce a relatively complex matrix of scatterplots with only one line of code. So, what does this pairs plot actually contain?

• The diagonal shows the names of the three numeric variables of our example data.
• The other cells of the plot matrix show a scatterplot (i.e. correlation plot) of each variable combination of our data frame. The middle graphic in the first row illustrates the correlation between x1 & x2; The right graph in the first row illustrates the correlation between x1 & x3; The left figure in the second row illustrates the correlation between x1 & x2 once more and so on…

In this first example, I have shown you the most basic usage of pairs in R. Let’s modify the options of the function a little bit…

## Example 2: Selecting Variables of pairs Plot

Often, you will only be interested in the correlations of a few of your variables. Fortunately, this can be done easily by specifying a formula within the pairs command:

`pairs(~ x1 + x2 + x3, data = data) # Produces same plot as in Example 1`

With the code above, we can create exactly the same plot as in Example 1. However, we can simply remove the variables from the formula, for which we don’t want to produce a scatterplot:

`pairs(~ x1 + x3, data = data) # Leave out one variable` Figure 2: Pairs Plot with Selection of Variables.

In this example, I deleted x2 from the formula, leading to a plot matrix that contains only the scatterplots of x1 and x3.

More modifications? Let’s do it!

## Example 3: Modify Color, Shape of Points, Labels & Title

In this example, I’m going to modify many different things:

```pairs(data[ , 1:3], col = "red", # Change color pch = 18, # Change shape of points labels = c("var1", "var2", "var3"), # Change labels of diagonal main = "This is a nice pairs plot in R") # Add a main title``` Figure 3: R Pairs Plot with Manual Color, Shape of Points, Labels, and Main Title.

The modified pairs plot has a different color, diamonds instead of points, user-defined labels, and our own main title. For even more options, have a look at the help documentation of pairs by typing ?pairs to the RStudio console.

## Example 4: Modify pairs R Plot by Group

Quite often you will have different subsets or subgroups in your data. Let’s add a group indicator (three groups 1, 2 & 3) to our example data to simulate such a situation:

```group <- NA group[data\$x1 < - 0.5] <- 1 group[data\$x1 >= - 0.5 & data\$x1 <= 0.5] <- 2 group[data\$x1 > 0.5] <- 3```

Now, let’s apply the pairs function again, but this time dependent on the group variable:

```pairs(data[ , 1:3], col = c("red", "cornflowerblue", "purple")[group], # Change color by group pch = c(8, 18, 1)[group], # Change points by group labels = c("var1", "var2", "var3"), main = "This is an even nicer pairs plot in R")``` Figure 4: pairs() Plot with Color & Points by Group.

As you can see in Figure 4, we colored the plots and changed the shape of our data points according to our groups.

You need even more options? No problem, let’s move on…

## Example 5: ggpairs R Function [ggplot2 & GGally]

So far, we have only used the pairs function that comes together with the base installation of R. However, the ggplot2 and GGally packages provide an even more advanced pairs function, which is called ggpairs(). Let’s install and load the packages:

```install.packages("ggplot2") # Packages need to be installed only once install.packages("GGally")   library("ggplot2") # Load ggplot2 package library("GGally") # Load GGally package```

The basic application of ggpairs is similar to the pairs function of base R. You simply have to write the following R code:

`ggpairs(data) # Apply ggpairs function` Figure 5: ggpairs R Plot via ggplot2 & GGally packages.

Main difference to the pairs function of base R: The diagonal consists of the densities of the three variables and the upper panels consist of the correlation coefficients between the variables.

Even better than pairs of base R, isn’t it? However, there is even more to explore. In case, you want to know more about the R ggpairs function, I can recommend the following YouTube video of the channel Dragonfly Statistics:

Please accept YouTube cookies to play this video. By accepting you will be accessing content from YouTube, a service provided by an external third party. If you accept this notice, your choice will be saved and the page will refresh.

Subscribe to the Statistics Globe Newsletter

• July 2, 2019 8:09 am

Hello Joachim, thanks for all your effort, this site is very helpful!
While trying to practice the pairs function along with grouping (specially example 4), I keep getting this error message:
Error in axis(side = side, at = at, labels = labels, …) :
invalid value specified for graphical parameter “pch”
Thanks so much

Thank you for the very nice feedback!

This error message typically occurs when the number of pch values is not the same as the number of groups.

In Example 4 we added this line to the code:

`pch = c(8, 18, 1)[group]`

With

`c(8, 18, 1)`

, we specified three different pch values for our three different groups.

If I would change the number of pch values (e.g.

`c(8, 18)`

), I would get the same error message as you.

• July 3, 2019 7:20 am

Cheers 🙂

• I’m glad that it helped 🙂

• Kevin Kranen
July 29, 2019 10:08 pm

Hi Joachim,
Great article. I’m running pairs() to correlate HVAC runtimes with power usage. I have set col=month where month is a factor that represents the month the data came from. Is there any way to either control the color for each month or plot a key in the base R version of pairs in this circumstance ?

• Hi Kevin,

Thank you very much for your comment. If I understand your problem correctly, Example 4 of this tutorial is what you are looking for. Your month variable would be the “group” variable that I have created in the example. Let me know whether you were able to fix your problem.

Regards,

Joachim

• Kevin Kranen
August 1, 2019 4:47 pm

Thanks Joachim,
That worked – I saw your approach earlier, but thought the group had to be numeric. Of course, factors work just as well.
Regards
Kevin

• Nice to hear 🙂

• Roberto León
December 20, 2019 6:49 pm

Joachim, than you, great site!!

I try ggpairs and got a nice graphics, however I also got a progress output about the grahph creation, fortunatelly, the function has a parameter to echo of: progress = F, here my script, where pariacaca_returns is a object xts.

ggpairs(as.data.frame(pariacaca_returns), progress = F)

• Hi Roberto,

Thank you for your nice words and also thank you for sharing your code!

Regards,

Joachim

• M Zam
November 9, 2020 10:52 am

How do i remove a column from my plot using pairs(data[, 1:7]). I need to remove column 2 from my plot as i do not need it

• Sheila
December 13, 2020 12:39 pm

Very helpful. Gave me a better understanding of the pairs function.

• That’s great to hear Sheila 🙂

• Nicolas
January 7, 2021 9:57 pm

Congratulations on the tutorial. It helped a lot. I had some problems with reproduction. I tried to manage the colors for different points or coordinates that meets my requirements but, I am not getting it. All of this using ggpairs.

• Hey Nicolas,

Thank you for the comment and the kind words!

In fact, my tutorial only explains how to color Base R pairs plots. However, I found this thread on Stack Overflow that explains how to color ggpairs plots as well.

I hope that helps!

Joachim

• Samina
January 14, 2021 8:59 pm

It really helped, I could generate fancy multivariate plot !

• That’s great Samina, glad it helped! 🙂

• Hollie
April 13, 2021 5:38 pm

Thank you so much! I had been struggling for a week on how to change the names down the diagonal of my matrix scatterplot and turns out it was so simple!

• Hi Hollie,

Thank you for the comment! I know this, you keep searching and searching and at the end it is just a little function or line of code that solves a problem… Glad it helped! 🙂

Regards

Joachim

• José
May 13, 2021 8:17 pm

Hi Joachim,

I have a lot of variables on my data set and while trying to plot them I keep getting a _figure margins too large_ error. How can I work around this? Is there a way to export pairs’ plot to a jpg or something like that?

• Wienche
June 7, 2021 5:20 pm

Hi Joachim, thanks for your explanations. Just a quick question is it possible to not get _all_ pairs but just the pairs between a bunch of independent vars and a bunch of dependent variables. I am not interested in the correlations between the independent vars among each other and the dependent variables among each other. So, I would like the correlations between X1 – X3 and X4 and X2 – X3 and X4 but not X1 – X2 and X3 – X4. I would like to have X1 and X2 on the x-axis of the grid and X3 and X4 on the y-axis. Is this possible?

• Hi Wienche,

Thanks a lot for the nice comment and the interesting question!

I had a quick look at the help documentation of the pairs function and haven’t found anything useful for this (not sure if I’m missing something here).

However, I would probably create such a plot using face_wrap/facet_grid of the ggplot2 package. You can find more info here: https://statisticsglobe.com/difference-between-facet_grid-and-facet_wrap-ggplot2-r

Let me know in case you have further questions.

Joachim