Multivariate literally mean “multiple variables“. That is, a situation where more than one factor varies, where those factors define for us, in some way, the situation. Bivariate pertains to two variable specifically and Univariate to one.
Lets consider an example in the world of websites: I’m building an E-Commerce website for a customer that sells gifts. In the design process there are a large number of factors that I need to consider before the final shop solidifies in to something I think is going to work well. These include (not exclusively) page width, font, key colours, logo position, right or left navigation, use of drop down menus, grid or list layout of products, etc, etc.
Some choices will be dictated by my client and some by best practice in usability and good taste, but that still leaves us with a lot of scope for variation.
I might decide to vary three element of just my home page and two landing pages that users are directed to from Google AdWords. For argument sake, lets say: page title size, layout style and colour of the “buy” and “checkout” buttons (choosing 10 different). To apply this I have a number of options:
Whichever, way I go I am going to want to measure how much revenue is generated over a period of time by each combination of variables. Lets say that there are 50 combinations of variables then each should be served and lead to a sale at least 50 times before we can get an significant result. That is 50*50=2500 sales. Great for a larger website, not so for a smaller site.
In other articles I will discuss this and other issues; explore some of the techniques that can be used; and take a broader look at website sales and sales conversion.
This is Multivariate analysis.