QuikSigma: Design of Experiments, Part 2

Today we go over Design of Experiments Part 2 in QuikSigma. Click here to go to Part 1.




Designed experiments are found here, under the Quik DOE link in the improve phase. When you click on that, you’ll get a page that looks like this. You’ll find a number of different designed experiments that you can do but just as we were doing in part one, we’re going to focus on a 2 to the K factorial experiment, and more specifically, we’re going to focus on full factorial designs where all possible combinations of input variables are done.

So the one that we looked at before was a three-factor eight-run and in this context these eight runs form what’s called a replicate, one kind of natural complete set of data. So all I have to do is click generate design, let me pop this up where you can see it a little better, and you may notice that we’ve got exactly the same balanced design that we had in the simulator. Low, high, low, high alternating and then two lows and two highs and then four lows and four highs and a place over here that we can record our output variable when we put these inputs into that state. Now, normally we will run with more than one replicate and to decide how many replicates we need. We come over here and we enter the smallest effect that we think will be of interest to us and the standard deviation of the process when it’s not being disturbed. Normally, we’ll just leave alpha at .05 and we see that our power is pretty low, small probability of catching an event if it happens. But as I increase replicates, my power meter comes up and when I get to, I don’t know, let’s try four replicates, now I’ve got pretty good power. 92 percent probability of catching the effect if it happens.

Normally, you will want to put the names of the variables in here and so let’s just call this one Time and let’s give it the same values that we saw in the in the simulator. 296 here for a low and 304 for a high and let’s give this, I don’t know, let’s call this one Pressure and give that the same values that we saw, which were if I remember correctly 50 and 110. C, I’m going to call, I don’t know, Catalyst maybe this is some kind of a chemical process and I’m going to get the same numbers that we had before 1.7 and 2.3. Now let’s see what we get. We should get 4 times 8 is 32 data occurring in natural blocks of eight. So let’s click our generate design and see what we get.

Ok, and expand, and sure enough I’ve got 32 and if you look, this pattern repeats. 1 through 8 are the first replicate, 9 through 16 are the second, and so on down and this is exactly what we generated and our rest simulator. Now, if I want to randomize my experiment, and that’s very good practice, that I just go to run order and click sort and ascending and now it’s randomized and if I want to put it back I just sort on design order and it will put it back the way it was. So pretty simple to get that set up and then we’ll go on to phase 3 and teach how to interpret an experiment.

Return to Blog Posts >>

Leave a Reply

Your email address will not be published. Required fields are marked *