Hi. This tutorial covers a specific type of experimental design called the randomized block design. So remember that any well-designed experiment tries to limit extraneous factors. Control, randomization, replication are all components of experimental design that help eliminate factors that could become confounding variables.
Blocking is another method. So let's define blocking. Blocking is sorting experimental units into similar groups called blocks using values of the extraneous factor as sorting criteria. So you want to define what your extraneous factor is, and then split that extraneous factor up into these blocks, these homogeneous groups, and then use those blocks within your experiment. Blocking in an experiment is pretty similar to stratifying a sample when you're dealing with an observational study.
So let's take a look at an example here. Suppose you're interested in studying how effective a certain adult flu vaccine is. A treatment group, which will be the vaccine group, and a control group, which will be the no vaccine group, will be used.
An extraneous factor that should be considered is the age of the participants. People of different ages have differences in exposure to the flu, differences in immune systems, and just health in general. Since age may confound the effectiveness of the vaccine, age should be the blocking variable.
So it's important here that you don't get all young people in the vaccine group and all old people in the no vaccine group. You're going to maybe get some biased results if that were to happen. So one way to eliminate that is blocking. So a possible blocking scheme that I'm going to propose is just to let block 1 be people aged 18 to 40, block 2 41 to 60, block 3 61 and older.
So when you're using blocking in the design of your experiment, generally you're going to want to consider using the randomized block design. And that's defined as a restricted randomization design in which the experimental units are first sorted into blocks, which are those homogeneous groups, so in our case, the h blocks, and treatments that are assigned at random within the blocks.
So if we go back to our flu vaccine experiment, so if the flu vaccine experiments were to use a randomized block design, a randomization procedure would be used to assign either the vaccine treatment or the no vaccine treatment to participants in each of the age blocks.
So if we think about our two treatments, vaccine and no vaccine, and then our three blocks, 18 to 40, so all of the 18 to 40-year-old participants are in that block, 41 to 60, and 61 and up, 61-plus, we want to make sure that we're going to assign each treatment to each block.
So one way to do this is that you could take one of the participants in this block, so one of the 18 to 40-year-olds, and just asked them to flip a coin. So if they get heads, they would receive the flu vaccine. If they got tails, they would receive the no vaccine. They would not receive the flu vaccine. So basically, what's happening here is we're going to be assigning both the vaccine and the no vaccine treatment to people within this block.
So by randomization, we're going to get some people with the vaccine, some people without the vaccine. And then we would do the same process with the 41 to 60-year-olds-- ask them to flip a coin. Heads, they get a vaccine. Tails, no vaccine. So again, we're assigning each of these treatments to people within each block. And then same thing for the 61-plus.
So using this randomized block design, we're getting people within each block receiving the vaccine and then also not receiving the vaccine. So that is an effective way of performing the randomization procedure using this block design.
Now, kind of one flaw of the randomized block design is that although randomized block design helps eliminate extraneous factors, each block would have a much smaller sample size than the original unblocked sample. So when the sample size is reduced, it's harder to reach definitive conclusions.
So if we go back to this experiment, we're going to have probably a pretty large sample here in our experiment. But if we break them into those three blocks, each of the block sample sizes will be much smaller than the original sample in general.
So we're not going to be able to come-- maybe be able to come to those same definitive conclusions that we would if we had all of these just as one group. So that is the tutorial on randomized block design. Thanks for watching.