In this quick lesson, we discuss how to use **QuikSigma****‘s** QuikScreen feature to better help you with your Design of Experiments. Have any questions about your own projects? Let us know below in the comments!

**Transcript:**

One very interesting and useful form of designed experiments is found here in QuikScreen. You’ll find that under screening designs and QuikScreen, or you can reach it under the QuikDOE panel. When you open that, that’s one of the available designs. You’ll see what we’ve got here is seven input variables A through G, and 1 replicate, consists of only eight ones. Normally when you do seven variables, if you do a full factorial, then you’re going to take a 128 observations. So eight 128s, turns out to be 1/16. This is a 1/16th fractional factorial and it is absolutely outstanding for isolating and finding the active variables and getting a rough model. It’s very quick and very effective. So as before, what you’re going to do is put all of your variables in their low state, take an observation. Then you’re going to put these three and a low state, and these four and a high state, take another observation and so on through.

Now, you’re going to end up doing this more than once during the course of the experiment. If I go down here to round 2 and click, you’ll notice that all of the variables change to red and all of the signs in the matrix flip sign. We start out with everything in its high state, then these three in the high state, and these four in the low state. So we’ve got basically a mirror image of the matrix we started with and that will give us some important clues as we go along. Then typically you’ll get to a round 3, and what you’ll do in round 3, is typically you’ll select one or two variables and you’ll flip those. So, and that sometimes

called folding, flipping the sign.

Now one of the things that the penalties that you pay for doing a 1/16th fractional factorial, is the confounding table is very long. You’ll have three interactions. Three two-way interactions confounded with each main effect. There are also three way, four way, five-way interactions, and the tables very long, but you’ll be right most all of the time if you assume that three-way interactions don’t happen. We just simplify and represent the two-way interactions. So where you use this is if you’ve got a set of input variables, you don’t know which is active, you’ll use this to find those and then fairly often you’ll do a full factorial on what remains. Let me just click advanced here and point out that were in the 8 run mode. There’s also a 16 run and that looks pretty much the same only twice as big and with twice as much confounding. So in this case, each main effect is confounded with, what’s that seven, two-way interactions, but the principle is exactly the same.