http://www.lunametrics.com/blog/2007/06/02/greg-niland-on-testing-not-so-excellent/ LunaMetric's blog on conversion rate and web analytics Tue, 07 Oct 2008 14:25:20 +0000 http://wordpress.org/ http://www.lunametrics.com/blog/2007/06/02/greg-niland-on-testing-not-so-excellent/#comment-1832 Matt Gershoff Thu, 24 Jul 2008 02:38:15 +0000 http://www.lunametrics.com/blog/2007/06/02/greg-niland-on-testing-not-so-excellent/#comment-1832 I think what he meant was fractional factorial design. Taguchi is just one way of setting up the experiment that takes advantage of prior information (assumptions) that the designer possess. The prior in this case is the belief that the factors under consideration do not all interact with one another. This means that the size of the space that needs to be explored can be reduced and you find your answer quicker. Of course if you are wrong about your assumptions then your results will be biased - something to consider when using MVT. I think what he meant was fractional factorial design. Taguchi is just one way of setting up the experiment that takes advantage of prior information (assumptions) that the designer possess. The prior in this case is the belief that the factors under consideration do not all interact with one another. This means that the size of the space that needs to be explored can be reduced and you find your answer quicker. Of course if you are wrong about your assumptions then your results will be biased - something to consider when using MVT.

]]>