Experimental design refers to how an experiment is carried out. Many experimental designs include a control group and a treatment group to compare effects of treatment (exercise, drug, video watching, etc.). You can have a good design of an experiment or a poor design of an experiment.
Good experimental design will have these three components:
Source: This work is adapted from Sophia author Jonathan Osters.
[MUSIC PLAYING] Hello. Let's take a look at a real-life example using experimental design. Suppose a farmer wants to try a new fertilizer in the fields. The three components of experimental design can be used to determine if the new fertilizer is better than the old one. Here's how it would work.
The first thing the farmer would do is determine the control by selecting 10 fields with similar soil nutrients, sunlight, and water. These are all variables that could affect the crop growth. The farmer would then apply the old fertilizer to five fields and the new fertilizer to the other five. By keeping the control elements consistent across the 10 fields, the differences between them can be isolated and attributed to either the old or the new fertilizer.
Next, the farmer takes randomization into account by randomly assigning which five fields will get the new fertilizer. While the fields selected were as similar as possible, there may be an unknown variable that was not accounted for. Perhaps some fields had moles underground. And that would affect how the crops grow.
By randomly assigning treatments, the farmer should get some fields with moles using the new fertilizer and some fields with moles using the old fertilizer. Randomization smooths out those effects that unknown variables might bring into the equation.
Lastly, the farmer understands the significance of repeated results rather than a one-off result. Say the farmer was only able to find two fields similar to each other and randomly assigned one for the new fertilizer and one for the old. It is possible in that case that the field with the old fertilizer does very well just by random chance. This would make it seem like the new fertilizer is not effective when perhaps it is.
Or the opposite could happen where it seems like the new fertilizer is effective when it's not. So it would always be better to randomly assign 10 fields as the farmer is more likely to find valid trends among 10 fields than two. Thanks for watching. And see you next time.
You could then apply the old fertilizer to five fields and the new fertilizer to the other five. By keeping all the other variables--soil nutrients, sunlight, water--consistent, the differences between the fields can be isolated and attributed to the old fertilizer or the new fertilizer.
Does the new fertilizer work? Is it effective? This is the idea behind controlling for all of these other variables.
EXAMPLEReferring to the farmer example, even though you made the fields as similar as possible with respect to water, sunlight, and soil, it's possible that there is a variable that you didn't think to control for. Perhaps some fields had moles under the ground, and that would affect how the crops grow. How would you know to control for moles?
Randomization in an experiment does not really achieve the same purpose as a random selection in a sample. When you do a simple random sample, the idea is to get a sample that's representative of the population. In an experiment, the purpose of randomly assigning individuals to groups is to filter out unknown sources of variation. The assignment in an experiment, however, is fairly similar to the way you would randomly select in a sample.
A larger size of the experiment means it's more likely that you can find trends that perhaps you wouldn't have found in a smaller experiment. The more you replicate, and the more experimental units you can get into your experiment, the more likely it is that you're going to find the true trends that arise, rather than some freak anomaly.