Source: Farm created by J. Gearin
This tutorial covers randomized block design-- a type of experiment. In randomized block design, first participants are divided into homogeneous groups-- groups that are the same across some variable of interest. Some examples are age, race, income, location, job, or gender.
Now, once the participants are in their similar groups, they are randomly assigned to treatment and control within that group. An advantage is that it controls for variables that would otherwise be confounding. If we think that job has an effect, this way we can make sure that a proportionate number of people who have the same job are assigned to treatment and control groups. A disadvantage is that it reduces the sample size for each group. Let's look at an example of what this might look like.
In this example, a researcher wants to study the effects of the two main types of laser eye surgery on participants. He thinks that age is a significant factor. So first the researcher would divide the participants into groups based on age.
These are some groups he could use. He could use under 30, 30 to 40, 40 to 50, 50 to 60, and 60 plus. Now, once people are sorted on age, you then randomly assign people within the group to one of the two treatments, so that that way there are a group of people under 30 who receive the first treatment, and who receive the second treatment. There are a group of people between 30 and 40 who receive each of the two types of treatments. It helps the researcher to examine what the effect of age is, and what kind of role that plays in how well the treatments work out for the people.
In this example, we have a farmer who's testing out fertilizer for a company. The company has three different types of fertilizer, and wants to know what's best. The farmer knows his land really well. He knows that it's not all the same.
He has a chunk by his house that gets a lot more sunlight, a chunk up by a river that gets a lot more water, and a chunk in the backfield that gets a lot more dry. If the farmer just applied the treatments randomly and had no concern for these differences of his land, the differences of the land would affect the results of the study.
If, for example, Treatment A got mostly in the water area, and Treatment A for some reason was really ineffective in a watery area, but overall was best everywhere else in his field, he wouldn't be able to determine that from the study. So instead, the farmer makes blocks out of his land.
He creates blocks based on what he knows to be similar. So he has a block down here for the really sunlit land, the block up here for the really watered land, and a block over here for the very dry land. Once he set up the blocks, he assigns the treatments randomly within each block.
Now, this is a very small-scale example. Obviously a company wouldn't just use one farmer. Most likely the farmer wouldn't assign just one set of block treatments. But it can give you an idea of what would happen.
In order to minimize the effects of the differences between the groups, we first make blocks, and then randomly assign the treatments within the block. This has been your tutorial on randomized block design.