Most scientific experiments are designed to test the effect of changing a single independent variable on a single or multiple dependent variables. However, sometimes the effect of simultaneously changing multiple interdependent input variables is required and this is called multivariate experimental design. For example, maybe doubling the amount of one chemical doubles the yield of a reaction but adding a catalyst makes doubling that one chemical triple the yield instead.

Multivariate Experiments

Instead of a list of every possible combination of variable settings, a Taguchi array (also called orthogonal array) can be used to create a minimal list of experiments to run. It can efficiently determine which factors are most influential on the results. For a more thorough explanation, watch NightHawkInLight’s video on the topic.

In their video, NightHawkInLight instructs viewers to search for arrays online and does not describe a way to generate your own. Searching online, I could not find a Taguchi array generator for multivariate experimental design so I decided to make one myself.

An algorithm for generating Taguchi arrays was surprising very difficult to find online but eventually, I came across a MATLAB script written by Chixin Xiao. I adapted it to JavaScript and created a web app.