Autoplot of PCA in R (Example)

8 Comments. Leave new

• 

  Alexandre

  May 16, 2023 6:33 pm

  Dear all, thank you very much for your collaborations.
  I have a not-so-trivial question.
  I would like to format the biplot variables in different ways. For example, the names Sepal.Width and Sepal.Length wanted the first to be Sepal subscribed Width and the other Sepal.Length in italics. Can someone help me?
  Very grateful for the attention.
  Yours sincerely,
  Alexandre Jardim.

  Reply
  • 

    Cansu (Statistics Globe)

    May 19, 2023 9:22 am

    Hello Alexandre,

    Renaming the variables is very easy using the names function. With that, you can rewrite Sepal.Width as Sepal_Width as follows.

    #rename data
    data(iris)
    names(iris)<-c("Sepal.Length", "Sepal_Width",  "Petal.Length", "Petal.Width",  "Species")

    Then you can perform PCA. As you said, labeling the vectors in italics is not trivial since autoplot() doesn’t provide a direct way to italicize loading labels. However, you can do it manually using the geom_segment and geom_text functions of ggplot2. See below.

    # load necessary libraries
    library(ggfortify)
    library(ggplot2)
     
    # perform PCA
    res.pca <- prcomp(iris[, -5], scale = TRUE)
     
    # get PCA plot for observations
    obs_plot <- autoplot(res.pca, loadings = FALSE)
    obs_plot
     
    # extract variable coordinates
    var.coord <- get_pca_var(res.pca)$coord
    var.coord
     
    # Add loadings to the plot
    final_plot <- obs_plot + 
      geom_segment(data = as.data.frame(var.coord), 
                   aes(x = 0, y = 0, xend = 0.15*Dim.1, yend = 0.15*Dim.2),
                   arrow = arrow(length = unit(0.15, "cm")), color="blue") +
      geom_text(data = as.data.frame(var.coord), 
                aes(x = 0.15*Dim.1, y = 0.15*Dim.2, label = rownames(var.coord)), 
                fontface = "italic", hjust = -0.1, vjust = 0.5)
     
    final_plot

    Please be aware that I rescaled the arrow lengths, which is done by default if you use the default settings of autoplot, to ensure that the arrows are visible in the graph. You can also try out other magnitudes to fit the labels better and use the just and just arguments for the adjustment. See the visual below.
     

    

    Regards,
    Cansu

    Reply
    • 

      Alexandre

      May 19, 2023 12:11 pm

      Hello Cansu,
      Thank you very much for your contribution.
      Unfortunately, my problem has been partially resolved. I would really like my variable to be subscripted (i.e., SepalWidth).
      I tried using the tag but it didn’t work. Do you know of any solutions?
      Regards,
      Alexandre.

      Reply
      • 

        Cansu (Statistics Globe)

        May 19, 2023 2:17 pm

        Hello Alexandre,

        You can easily change it by using the names function as I previously shared names(data)<-c("anyformat",...). If it doesn't answer your question, could you please give some examples for me to understand? Regards, Cansu

        Reply
• 

  Alexandre

  May 19, 2023 7:03 pm

  Dear Cansu,
  I am very grateful that you have gone to great lengths to help me with this question.
  Unfortunately, I still can’t solve my problem.
  I am sending an image as I would like the variable names to be in the biplot.
  https://i.im.ge/2023/05/20/hTLy8Y.Ex.png
  Thanks again for all your help.
  I hope you can help me.
  Yours sincerely,
  Alexandre.

  Reply
  • 

    Cansu (Statistics Globe)

    May 22, 2023 8:16 am

    Hello Alexander,

    Thank you for the sample image. Creating a label variable using square brackets could help as follows. See the tutorial: Add Subscript and Superscript to Plot in R for further details. The general structure of the script is the same.

    # load necessary libraries
    library(ggfortify)
    library(ggplot2)
     
    # perform PCA
    data(iris)
    #rename data
    data(iris)
    names(iris)<-c("Sepal.Length", "Sepal_Width",  "Petal.Length", "Petal.Width",  "Species")
    res.pca <- prcomp(iris[, -5], scale = TRUE)
     
    # get PCA plot for observations
    obs_plot <- autoplot(res.pca, loadings = FALSE)
     
    # extract variable coordinates
    var.coord <- get_pca_var(res.pca)$coord
     
    # Create new labels
    # Create new labels with subscript
    new_labels <- c("Sepal.Width" = "Sepal[Width]", 
                    "Sepal.Length" = "Sepal[Length]", 
                    "Petal.Width" = "Petal[Width]", 
                    "Petal.Length" = "Petal[Length]")
     
    df<-as.data.frame(var.coord)
     
    df$labels <- new_labels
     
    # Add loadings to the plot
    final_plot <- obs_plot + 
      geom_segment(data = df, 
                   aes(x = 0, y = 0, xend = 0.1*Dim.1, yend = 0.1*Dim.2), 
                   arrow = arrow(length = unit(0.2, "cm")), 
                   color = "black", 
                   inherit.aes = FALSE) +
      geom_text(data = df, 
                aes(x = 0.1*Dim.1, y = 0.1*Dim.2, label = labels), 
                parse = TRUE,  # parse mathematical expressions
                fontface = "italic", hjust = -0.1, vjust = 0.5, 
                inherit.aes = FALSE)
    print(final_plot)

    

    Regards,
    Cansu

    Reply
• 

  Alexandre

  May 22, 2023 2:36 pm

  Hi Cansu,
  Thank you very much for your efficient response and your time dedicated to solving it.
  My question was literally solved!
  The work you do is fantastic, thank you very much.
  All the best.
  Regards,
  Alexandre.

  Reply
  • 

    Cansu (Statistics Globe)

    May 22, 2023 2:53 pm

    Hey Alexandre!

    Glad to hear that I could help!

    Best,
    Cansu

    Reply