What is Multivarable Testing?
September 19, 2006 10:28 AM Subscribe
What is Multivariable Testing, really? It was mentioned in this NYTimes article, but the term is only ever used in relation to QualPro, Inc.. What is the real mathematical/statistical technique they're using? Is this related to the Taguchi Method? Can someone explain it to me?
To clarify (perhaps needlessly), I think the point is that during testing you can change many variables simultaneously. This speeds your work because you don't have to go through every single potential combination of variables - a reasonable subset will do. Having done this, you need to do something special to glean information out of that information; usually, some regression technique that will pick apart the effect of each variable in the big mash that you tested.
posted by metaculpa at 2:07 PM on September 19, 2006
posted by metaculpa at 2:07 PM on September 19, 2006
To add to what metaculpa has said, there are many ways of doing analysis of multiple linked variables. Multivariable Testing™ is just one implementation, so don't think that they've got the one and only special voodoo for doing things like were mentioned in the article. Bayesian methods are all the rage now in scientific analysis, and they'll probably spill over into the less theoretical fields soon, if they haven't already.
posted by Mr. Gunn at 7:58 PM on September 19, 2006
posted by Mr. Gunn at 7:58 PM on September 19, 2006
« Older What US law schools have programs in either... | Looking for simple content management for... Newer »
This thread is closed to new comments.
My personal guess is that it's some derivate of partial least squares regression, which is a useful way of finding the significant correlations between predictor and response variables. But that's only because it's my favorite technique that no-one's heard of.
posted by metaculpa at 10:54 AM on September 19, 2006