rugplot: PCA Projections#

A popular technique to visualize the result of Principal Component Analysis (PCA) is by means of projections where data with m columns or variables are projected into a subpace with fewer columns, typically 2. Here it is shown how to create PCA projections using the rugplot container.

The famous iris dataset, (Fisher, 1936) will be used to create such examples. The dataset can be downloaded from different sources and in different ways, three of them are: first, visiting DataHub, second by running the following command

wget https://datahub.io/machine-learning/iris/r/iris.csv

and third, by adding the link in the previous command to the JSON template that will be created in the second step for creating a PCA projection.

Creating a PCA projection plot using the rugplot container#

For simplicity it is better to create an alias, see the Docker commands section.

  1. Step 1, create a rugplot PCA projection template

    rugplot template --plot pca

    A pca_projection_params.json file will be created including some of the name/value pairs listed below:

    {
    "description": "Parameters for a PCA projection using the 'rugplot' R package",
    "filename": "<filename path>",
    "colour": null,
    "labels": {
        "title": null,
    },
    "save": {
        "save": false,
        "outputfilename": null,
        "device": "pdf",
    }
}

  2. Step 2, add the 'data file', 'colour' and the 'title' values in the template:

    {
    "filename": "https://datahub.io/machine-learning/iris/r/iris.csv",
    "colour": "class",
    "labels": {
        "title": "Iris PCA projection",
    },
    "save": {
        "save": true,
        "outputfilename": "iris-pca.png",
        "device": "png",
    }
}

  3. Step 3, create the PCA projection plot

    rugplot plot --plot pca --file pca_projection_params.json

    The result will be stored in the Rplots.pdf file.

    pca projection result

Interactive plots using plotly

Interactive plot#
{
    "labels": {
        "title": "Interactive iris PCA projection",
    },
    "save": {
        "outputfilename": "iris-pca.html",
        "device": "html",
    }
}

The result