I am doing some research on how to analyse experimental data. During the last three weeks, I face the expression "experimental design", I try to google to understand this notion, and how it works.

What I understand is that it is a way to analyse experimental data using designs . By design it mean any form (linear and even nonlinear).

However, till now, I haven't feel i am understanding this concept correctly, would some one please lead me to some good and simple references on such subject? Or would some one give me a brief explanation of what is experimental design , and what is the difference between "experimental design" and "design of experiment" ? Thank you in advance.

  • 1
    $\begingroup$ I took two courses in design of experiments in a grad program in statistics, and I am very, very far from being an expert in the area. I am amazed that you seriously intending to learn this by googling the phrase. $\endgroup$ – StasK Jan 28 '16 at 16:03
  • 1
    $\begingroup$ @StasK Sorry, but May I ask you way ? If you think this is not the optimal way to investigate such topic, I would like to hear from you other ways I can follow. $\endgroup$ – Nizar Jan 28 '16 at 16:10

There is likely no difference between "experimental design" and "design of experiments" except perhaps "design of experiments" applies more theoretically and generally to all experiments while "experimental design" usually refers to the design of one particular experiment. However there is no substantive difference between the two, and you could use the two interchangeably to refer to the same thing.

Experimental design is, ultimately, about how your experiment is constructed. If you want to test the effects of a drug, you may split people into two groups and administer the drug to one of the groups. You may decide instead to have three groups, where one group gets the drug under investigation, one group gets a drug that is already in the marketplace, and one group gets nothing. This would be a difference in "experimental design" because you are deciding how many groups you want to include.

When splitting people into groups, you may consider certain variables (like age or gender) for which you wish to control, so that means you might make sure that each group is 50% men and 50% women. This is a design choice - do you want to control for age/gender/other variables?

Sample size is also an aspect of experimental design. If you split people up into groups, should the groups be equally sized? Is there a reason, due to cost or availability of the drug, that your group reviving the drug can only be so big, so the group receiving the drug and the group not receiving the drug may be of different sizes?

I've briefly mentioned a few basic aspects of experimental design here but entire areas of study are devoted to the topic, so sample size, number of groups, and whether or not to control for certain variables are certainly not an exhaustive list. What I recommend is to answer the question "What is my goal here?" Depending on your goal (i.e. see if a drug has a particular effect on people), I would use that as the starting point and develop the simplest experimental design that allows you to completely answer your question.

| cite | improve this answer | |
  • $\begingroup$ Thank you vary much for your answer, however I am still a little bit confused. Indeed your answer gives a good example that explain how we do an experimental desing. However I am searching for " How can we explain (theoretically) what is an experimental desing for some one out of the domain" ? $\endgroup$ – Nizar Jan 28 '16 at 14:41
  • $\begingroup$ What do you mean by saying "for someone out of the domain?" $\endgroup$ – Matt Brems Jan 28 '16 at 14:45
  • $\begingroup$ Thanks for replying, I mean some one who is not familiar with data anlysis. In fact I mean "me" . Indeed it seems very difficult for me to understand what is DOE and why we use it ? we apply it before doing an experiment or after ? Its main goal is just to organize the way we use to do the experiment ? we construct such design before collecting data or after ? Sorry for my odd questions. $\endgroup$ – Nizar Jan 28 '16 at 14:52
  • 1
    $\begingroup$ Experiments are a method of data collection. You design your experiment (your data collection) in a way that makes the most sense and allows you to get the most out of your analysis. So you'll design your experiment ahead of time in order to collect your data properly. $\endgroup$ – Matt Brems Jan 28 '16 at 14:55
  • 1
    $\begingroup$ Approximately, yes. Whenever you conduct an experiment, you are creating a design for the experiment. You may not think about it, but you are. When you pick 20 students in a science class and split them into two groups of ten, that's a particular experimental design. Whenever you conduct an experiment, there is always some underlying design to it. Your choice of design (sample size, factorial design, etc.) should be motivated by what is sufficient to answer your question. $\endgroup$ – Matt Brems Jan 28 '16 at 15:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.