How to Approach Design of Experiments in a Practical Way
A Scientific Method for Experimentation
How to Approach Design of Experiments in a Practical Way.
[leadplayer_vid id=”52594F7A0148B”]
Lots more to come. Subscribe.
SLIDE: A Scientific Method Model for Experimentation
So here is the scientific method model for experimentation. It’s a little bit messy graph, but I hope you get the point. We’ll refer to this map as we go along as well. First, your questions, your hypothesis, will have to come out of the Business Need, as it’s worth your time and resources. Some questions you can just perhaps go over without answering, if it not in financial need, and then when you form that question, or when you form that hypothesis it has to be of practical, not statistical, okay. So do not proceed until practical difference has been defined, and then when you plan your study it’ll be in iterations. So don’t spend your entire budget in the first…the rule of thumb is twenty-five percent on the first experiment of your budget, so forth, cause you’ll have follow-up experiments, the answers that have not been…or the new questions that will be raised from your first experiment, so forth. So investigation is iterative in nature, or data from the first experiment are comparative predictions, which is from your hypothesis. So, always make predictions and you’ll see how we do this in our jump, and so forth.
Now, the prediction…the reason this is important, predictions is equals…when you get data after your predictions…the difference between your predictions versus the actual data…actual difference equals you knowledge gap. So, if you knew what your process…the result of your process, then you have very little knowledge gap. That means you have good understanding of your process. However, if there is a big gap, the delta, between your data and your prediction, then you have a huge knowledge gap. So, that’s kind of the metric of how well you understand your process. And, then we go into the Run Study, this is where you get data, and so forth. Finally, we draw conclusions, so this is where the predictions versus data, have the study met objectives? What relationships need more explanation? New questions come up, we plan, and we go into the next study, and so forth, new knowledge, and we’ll have to go back into our process map, hypothesis map, measurement systems, (analyses and so forth). So, it’s all iterative, so you have to take experiments in small chunks, don’t try to plan one massive, giant experiment that will consume all your resources in time, because you did not at the time…you may be at the point where you do not fully understand what you need to experiment on.
Lots more to come. Subscribe to learn more!