Experiments 4E - All about blocking скачать в хорошем качестве

Experiments 4E - All about blocking

Videos used in the Coursera course: Experimentation for Improvement. Join the course for FREE at https://www.coursera.org/learn/experi... These videos are also part of the free online book, "Process Improvement using Data", http://yint.org/pid Full script for the video: http://yint.org/scripts/4E -------------------- This class is all about blocking. The distinction between covariates and disturbances is in the prior class, was that covariates are measurable, and disturbances are not. Neither of them can be controlled. We're generally comfortable with the idea of control in our experiments. Variables that we can control, should be controlled. Remember that idea that you should always keep everything fixed, and only change what you're experimenting with? The idea of keeping things fixed is what we mean by the term "control". Now we're going to make a subtle distinction. There are variables you can control, and choose to actively vary. We call them "factors". You can control these variables, and you can measure them. Remember the cell phone app example from the previous class? Recall the variables from that example were : Factor A was the type of marketing promotion used. Either a free in-app purchase, or a 30 day trial of all the features. Factor B was the marketing message and Factor C was the in-app purchase price. All three of those were controlled and measurable where the term "measurable" is used fairly loosely as a way to say that you can quantify the value of your factor. Take a minute now, to think about the experiments, and the factors you have varied, during this course. Are your factors actually controllable? Can you measure or quantify all your factors? If you have any doubt, it's a good time to consult with your colleagues and post a message in the course forums. Now here's the subtle distinction. What if you had a factor you can control, it varies during the experiments, but the factor isn't actually the main focus of your experiments? These are called nuisance factors. Let's take a look at the cell phone example again. The end user could be on an Apple or Android operating system. You could control this, because you can select only Apple or Android users during your experiments. In fact, this could have been another factor. You might have called it Factor D, but this really isn't expected to be a significant factor of interest. It's not what the aim of your experiments are about. We call this a "nuisance factor". Now from the previous video, you learned that you must randomize your experiments. If you go run all your Apple experiments first, and then all your Android experiments after, you've actually confounded your variables on this factor D. Ultimately, you would want successful sales, no matter whether you have Apple or Android users, but you must intentionally plan your experiments ahead of time, to avoid this nuisance factor from having a confounding effect. We call this "blocking". There are a number of instances in which you want to block for nuisance variables. In our baking example, imagine a situation where you are going to run out of flour. You can do half your experiments on one brand of flour, and half the experiments with another brand of flour. Another case might be experiments in a factory. Half the experiments are done during the day shift, and the other half is done during the night shift. Or if you're testing gas mileage in a car, you might have one driver and another driver. People sometimes find blocking tricky to understand. And here's one way I deal with it. I ask myself, is my system, or process, going to have to successfully work with different values of this nuisance variable in the future? If the answer is yes, then I design with blocking in mind. If I don't design my experiments with blocking in mind, then that nuisance variable might affect the outcome, and I won't really understand what's happened. If the answer to the question is "no", that means that I've got good control over the system, and I can avoid the nuisance variable. Now that we've discussed the need for blocking, let's see how to plan experiments where there are 2 blocks. And the general rule for this situation is to add a new factor to your standard order table. We already have 3 factors, A, B, and C. So that's 8 experiments in a full factorial. Now consider adding the new nuisance factor as a new variable, D, to the table: Apple versus Android. This factor has two levels; minus and plus. How do we go about picking which experiments to assign to our Apple phone users, and which experiments to assign to our Android users? Here's a hint. If we were to do a full set of experiments in four factors, we would've required two to the power four, or 16, experiments. Instead, we're doing eight experiments shown here. Eight is half of sixteen, and so there's no surprise that this is effectively generated using a half fraction. Once you understand the principle of ...