The randomized block design is equivalent to the stratified random sampling in research designs. A block is a group of experiments subjects that are known to be somehow similar before conducting the experiment and the way in which they are similar is expected to have an effect on the response to the treatments. Like stratified sampling, the key purpose of randomized block design is to reduce noise or variance in the data. Generally, researchers should group the samples into relatively homogeneous subunits or blocks first. Then the random assignment of subunits to each treatment is conducted separately within each block. Since the variability within each block is less than the variability of the entire sample, it is more efficient to estimate the treatment effect within a block than across the entire sample. So, if we use these more efficient data across blocks, we should get a more efficient estimation than we would without blocking.
One quick example: assuming that we are conducting a simple posttest-only randomized experimental design during which we find that our samples have some intact or homogeneous subunits. To be specific, in a research of high school students, we expect that they are relatively homogeneous with regard to class or year, which means we hypothesize that the variability within grade is less than the variability for the entire school. Then we may block them into 4 units namely freshman, sophomore, junior, and senior. If our hypothesis is correct, we can get more effective estimations of the treatment effect within each block and then conduct the simple posttest-only randomized experiment within these four blocks.
Here are what you should know about randomized block design:
1. As an external observer, you may not know that you are blocking and you are implemented the same design in each block. Because people in each block are not segregated or separated from each other. That is to say, although your research participants are blocked, there is nothing affected by this design strategy (Figure 1). Actually, this analysis strategy makes your data have fewer noises when you assign your participants into different blocks.
Figure 1. Three treatments are arranged in two blocks (k=3). Each block is divided into k sub-blocks, where k is the number of treatments. And each block receives all the treatments that are randomly assigned to the sub-blocks.
2. Only if your hypothesis is right, that is, the variability within each block is less than that of the entire sample, can a blocking design contribute to your research. If classes within different grades are not relatively homogeneous with regard to your measures, you may get no benefits from block design and make an actually less effective estimation of the treatment effect.
So, how do we know if a randomized block design is appropriate? You should first figure out whether the units are relatively homogeneous. In the example we mentioned above, if you are studying how long they read every day, is it appropriate to assume that freshmen are relatively homogeneous to each other than they are to juniors or seniors? Would they be more like each other with regard to measures related to vocabulary size?
3. Your decision on block affects the judgment on the part of the researcher.
At CD BioSciences, we can help you not only the randomized block design but also the appropriate choice of statistical strategy. If you have any questions, please feel free to contact us.