Simon Potter

© 2024 Simon Potter CC BY-SA 4.0

Masters Thesis

During my Masters year I worked on a thesis entitled “Dynamic, Interactive and Reactive Statistical Graphics for the Web”. The product of my thesis is a cohesive link between the powerful interactive and graphical capabilities of a web browser, and the statistical graphics and computing engine provided by R. My thesis is available on ResearchSpace or on this website by clicking the thumbnail below:

Masters Thesis

The basic idea is that while R is great for statistics, its graphics engine does not support dynamic and interactive graphics such as those created using D3. Additionally, while D3 is great for dynamic and interactive graphics, we have to get statistical information from elsewhere to visualise anything more than the most basic of statistical models. To get the best of both approaches, I further developed the gridSVG package to be a bridge between the graphics provided by R and the interactivity and animation provided by a web browser.

To cite this thesis, use the following BibTeX entry:

@MastersThesis{SJP2013,
    author = {Simon Potter},
    title  = {Dynamic, Interactive and Reactive Statistical Graphics for the Web},
    school = {The University of Auckland},
    year   = 2013
}

Seminar

On the I gave a seminar presentation at the University of Auckland in the Department of Statistics. The slides can be viewed by clicking the thumbnail below:

Seminar

Some of the examples in the seminar were R web applications that were running from my laptop. To view these examples, you will need to use the sjpMScThesis package, which is available later in this page. The examples I used were the following:

  1. loessStatic on slide 11.
  2. loess (or loessShiny, the effect is the same) on slide 34.
  3. arima on slide 44.
  4. sampvarClean on slide 54.

Examples

Much of the content demonstrated in my Masters Thesis displays animated and interactive content that are not adequately captured in printed form. Furthermore, many of the examples also demonstrate the use of R as a web server, which is central to the core idea of my thesis.

I have made the examples shown in my thesis (and others) available as an R package. This package is called sjpMScThesis, and the examples within the package depend on a few packages being available on an R installation. The following R command will install some of these packages:

install.packages(c("gridSVG", "ggplot2", "selectr",
                   "hexbin", "shiny", "Rook"))

There are further packages that require installation that are not (yet) available on CRAN. These packages are animaker and the companion package to my thesis sjpMScThesis. These are available in the Downloads section below.

To view videos of examples created as part of my thesis, click one of the links below:

  • Hexagonal binning — A comparison of encoding the density of data in a plot by area or colour.
  • ARIMA Diagnostics — Interactive ARIMA model diagnostics. ACF and PACF plots dynamically update in response to changes in any of the p, d or q parameters in an ARIMA model.
  • LOESS Smoothing — A simple example which easily enables a user to determine a suitable value for controlling the degree of smoothing applied.
  • Multidimensional Scaling — An interactive example showing the effect of multidimensional scaling on the true pairwise distances between european cities.
  • Sampling Variation Teaching Example — A reimplementation of the sampling variation animation that is created by the Visual Inference Tools package. Key improvements are that animation timing is accurate and the animations themselves are smooth.

Download

animaker

animaker is developed on GitHub and is not (yet) available on CRAN. The following links are simply builds from the latest revision of the GitHub repository.

sjpMScThesis

The sjpMScThesis package is developed on GitHub. The following links, like animaker, are builds from the latest revision of the GitHub repository.

Usage

Each of the examples in the sjpMScThesis package are designed not only to be easy to use, but also to be easily modified or examined. If you inspect the examples folder where the package was installed, you will find each of the examples. They should all run outside of the package with minimal modification (only relating to source() usage and Rook port selection).

To run the “LOESS Smoother Example”, all that is required is the following code:

library(sjpMScThesis)
thesisExample("loess")

To stop an example, simply call stopThesisExample().

For a list of the available examples, see listThesisExamples().

> listThesisExamples()
 arima: Interactive ARIMA Model Diagnostics
 hexbin: Hexagonal Binning Comparisons
 hexbinClean: 'hexbin' with no explanatory paragraphs
 loess: Interactive LOESS Smoothing
 ...

Examples can be configured using the getExampleOptions() and setExampleOptions() functions. Consult the sjpMScThesis package documentation via ?setExampleOptions for more information.