a web-based analysis of variance

anova-web is a web based suite of small utilities that I use as the backbone of a Master's Course on Experimental Design, a mandatory course for the Masters in Biodiversity, Genetics and Evolution, and an optional course for many other Masters held at the Department of Biology of Faculty of Sciences, University of Porto.

The course is loosely based on A. J. Underwood's book Experiments in Ecology: Their Logical Design and Interpretation Using Analysis of Variance, Cambridge University Press, 1997 ISBN: 978-0521556965. The focus is put on hypothesis testing and how to devise experiments that can answer unambiguously questions pertaining to the highly variable biological world and its phenomena.

Initially, the software for the course consisted on a web site written in PHP which used a CGI executable written in C++ to compute ANOVAs. This software is still available at https://github.com/sairum/mwanova. However, due to security reasons the IT department strongly encouraged me to abandon CGI executables and use instead ECMAScript (JavaScript). Hence, anova-web is precisely the port of the old C++ mwanova to a web-based software.

While porting anova-web I aimed at developing a suit of software with the least amount of dependencies possible. Using node.js was my first attempt at porting anova-web to a JavaScript environment. I used grunt as the task runner of choice, but the incomprehensible large amount of node modules installed soon made me abandon node as a developing platform and use plain JavaScript. Abandoning node implied abandoning grunt and all the 'minifications' and linting facilities that come with node. Currently, anova-web uses a plain Makefile to build all its components. If you want to use it just type

$ make watch

and edit the files under inside the folder src/ (this includes css styles and JavaScript). To make use of the Makefile you need inotifywatch, a standard tool installed by default in most Linux system, and Douglas Crockford's jsmin (https://www.crockford.com/jsmin.html) for minification of JavaScript files. Other than that, anova-web depends only on three JavaScript libraries and one CSS Style sheet:

• JStat (https://github.com/jstat/jstat) for probability distributions and related stuff
• FileSaver.js (https://github.com/eligrey/FileSaver.js/) to generate files to download
• seedrandom.js (https://github.com/davidbau/seedrandom) as a RNG with seed capabilities to be used by JStat
• normalize.css (https://necolas.github.io/normalize.css/) to make browsers behave well

The proper way to install anova-web is to clone this project and copy all *.html files, together with the css/, examples/, js/ and support/ folders to the document root directory of an http server such as apache:

• css/
• examples/
• js/
• support/
• index.html
• anova.html
• examples.html
• sampling.html
• simulate.html

The src/ directory is not necessary unless you intend to modify or improve the software. All code resides on the js/ directory, either minified or not. The *.html files use the minified versions of the JavaScript files found under js/.

Another way to run anova-web is to uncompress or clone this github repository somewhere in the directory tree of a personal computer. Then, using a GUI, click on index.html. This should run the whole bundle just fine (at least it does on a Linux system running KDE Plasma 5).

openDataFile()
 │
 ├── resetAnalysis()
 │
 ├── DEBUG: displayFactors()
 │
 ├── displayData()
 │
 └── computeCells()
       │
       ├── DEBUG: displayCells()
       │
       ├── homogeneityTests()
       │
       ├── setupPostHocTests()
       │
       └── buildTerms()
             │
             └── IF ── getCellsSS()
                  │
                  ├── DEBUG: displayTerms()
                  │
                  ├── correctForNesting()
                  │     │
                  │     └ correctTermNames()
                  │
                  ├── DEBUG: displayFactors()
                  │
                  ├── computeCTRules()
                  │     │
                  │     └ displayCTRules()
                  │
                  ├── DEBUG: displayTerms()
                  │
                  ├── displayAverages()
                  │
                  ├── buildPostHocTests()
                  │     │
                  │     └── computePostHocTests()
                  │
                  └── displayANOVA()