Design of Experiemnts
Wikipedia probably explains it better, but I'll try to summarize.

One of the posters on the previous page suggested that in order to understand the influence each factor has on the finished product, I should hold all factors constant except one, which I vary by known quantities, and then I can observer the outcome of that factor. Then I repeat ad nauseum for each factor until I fully understand how all of these variations result in different tasting beers.

The main idea with Design of Experiments is that you set up a series or set of experiments to account for not only the effects of single factors, but the interaction effects between multiple factors. So in the case of the dry-hopping experiment, let's say I have the following factors I'm considering:

Hop Variety

Amount of Hops

Length of addition

Time of addition

By setting up a properly designed experiment, I can also observe the interaction effects between Amount & Time, Length & Time, and if I wanted to run enough levels, even Amount of Hops, Length, and Times of additions.

When I mentioned Full Factorial before basically that just means running all the combinations between the factors. In this case with four factors I'm looking at a minimum of 16 combinations to test. That is, if I do two "levels" which means I only pick two values to try for each factor, I end up with 2^4 design points. However, using two levels assumes that there's a linear effect on the results by all the factors. E.g. Adding 1oz. of Columbus is going to leave twice the impression as 0.5oz. I would say that's unlikely, so I'd need to do at least 3 levels which would allow my to fit a quadratic curve to the results. However 3 levels means you'd have to perform 3^4 = 81 tests for a Full Factorial design.

Hence why I mentioned other types of designs like Plackett-Burman, which allow you reduce the number of experiments and still observe interaction effects.