4/5 (1)

If I would ask someone that is not focused in the area of analytics about his idea of a design, the answer that I would get is probably something like the following

“A design is what gives things the color, shape, and form.”

However, when an analyst talks about design, he talks about how he will or has collected data. Depending on the purpose, the design might include more information than the way one collected the data.Let me give you an example, when the analyst explains the experimental design, he will also start discussing about the orthogonal design, fractional factorial design or the treatment groups.

Analytical designs are the fundamentals of logic and cause-detection. Not only do they help the analyst collect data in a systematic way, but they also make it possible for the analyst to draw a logical conclusion based on the idea of ruling out alternatives. For a strong argument to be made, it is necessary to rule out all the alternative explanations.

At the end of the day, it’s not the fancy statistical methods that enable each analyst to draw logical conclusions. Why do I say this? Well, imagine reading an article that talks about a correlation found between ice cream consumption and the number of people drowned and the causality that exists between the two (I hope you won’t read such an article although in today’s world anything can happen). You, being the analyst you are, should take this article and invalidate its findings for the simple reason that the data was not collected in a systemic way. By this I mean that the design does not allow “the author” of this article to rule out alternative explanations that might exist to explain the correlation.

(By the way, the most common sense answer to the following example fellows would be that we find a high correlation because people consume more ice cream in summer and drownings happen more in summer. In other words it is just a spurious relationship).

In this section, I will show several ways to apply different designs, their advantages, disadvantages, and limitations. I will cover not only the experimental designs and the popular AB-test, but also a wide range of other types of designs including here quasi-experimental designs, sampling questions, the logic of causal inference, qualitative approaches like case study designs, survey designs, and mixed-methods designs. So jump on board!

Give me feedback, and tell me how you liked the article ;)!