What happens when a researcher has many groups in their study? In this lesson, we’ll look closer at multiple-group design, including multiple-group design with independent groups and multiple-group design with correlated groups.
Shakira is a psychologist who has been hired by an advertising agency. She needs to know what type of spokesperson works best to sell products to teenagers. Do ordinary people work better than celebrities? Should it be a teen or a parental figure? Or should companies just sell their products without a spokesperson at all?Shakira has to decide how to answer those questions in an experiment. The process of making decisions about what type of experiment to run is called experimental design. There are many different ways that Shakira can organize her study and many different decisions she has to make as she is designing it. Let’s work with Shakira to see what decisions she has to make and to look closer at one type of experimental design, the multiple-group design.
Levels of Variable
So, Shakira wants to know what type of spokesperson makes teenagers want to buy products. She decides that she’ll gather teenagers and show them one of several different commercials for a product, and then let them decide if they want to buy the product or not. The commercial that results in subjects deciding to buy the product the most is the one that works best.Pretty simple, right? But when Shakira goes to make the commercials for the study, she realizes she has a lot of spokesperson options: she has to make a commercial without any spokesperson, one with a non-famous person, one with a celebrity, and one with a parental figure. The levels of a variable are the different options of that variable. In Shakira’s case, her independent variable (the spokesperson) has four different levels: none, non-famous, celebrity, and parental figure. Her dependent variable (sales) has two levels: the subjects can buy or not buy the product.
Think about the variable gender; it has two levels: male and female. Someone who is studying age might create three levels for that variable: young, middle-aged, and old. If letter grade is your variable, you might have five levels: A, B, C, D, and F. The point is that every variable has levels. Some have more than others, which brings us to the idea behind multiple-group designs.
Now, Shakira has a complicated study with four different levels of her independent variable. But what if she had a simpler design? What if she just wanted to know if having a spokesperson would make teens more likely to buy? In that case, she only has to make two commercials: one with a spokesperson and one without a spokesperson.A two-group design in experiments involves having only two levels of your independent variable. Usually, this means a yes or no situation: yes, the commercial has a spokesperson, or no, it does not.
But what about with Shakira’s study with the four different levels? If an independent variable has more than two levels, it is called a multiple-group design. Usually, one of the levels is nothing at all and the other levels are variants. For example, Shakira has a no-spokesperson group, and then she has three different variants of a spokesperson: non-famous, famous, and parental.
Two-group and multiple-group designs are both valuable, but they answer different questions. A two-group design tells you whether your independent variable has an effect at all, while a multiple-group design tells you how much of an effect each level has. For example, if Shakira only wants to know if a spokesperson increases sales, she can go for a two-group design.
But if she wants to know what level (or type) of spokesperson increases sales the most, she will opt for a multiple-group design.
Let’s go back to Shakira’s original study. She wants to know which level of spokesperson has the biggest effect on sales, so she chooses a multiple-group design. That’s it, right? She’s ready to run her experiment! Well, not quite. She still has to decide which type of multiple-group design she wants to use. That’s right; to make this more complicated, there are multiple types of multiple-group designs. But don’t worry; we’ll break each of them down.
Essentially, the type of multiple-group design that Shakira chooses is based on how she assigns subjects to each group. Which subjects see the celebrity spokesperson? Which ones see the parental one? Ideally, she wants her groups to be equivalent. That is, she doesn’t want all the people in one group to have a variable in common that affects sales.
For example, imagine that Shakira has 100 subjects and she assigns 25 of them to each of her four groups. But when she does that, she ends up with all the shopaholics in the group that watches the commercial for the celebrity. Meanwhile, all of her misers end up in the group with the parental spokesperson.When it’s time to buy or not buy the product, Shakira can’t know if a celebrity spokesperson works better than a parental one or if that group bought the product more often because they had a bunch of shopaholics in it and the parental group didn’t. But if the shopaholics and misers are equally spread out across all four groups, then she has four equivalent groups. There are many ways for Shakira to get equivalent groups, and they are divided into two types of multiple-subjects design.
A multiple-subjects design with independent groups is when Shakira randomly assigns subjects to a group. She could flip a coin or draw names out of a hat or use an online randomizer tool. As long as she is truly randomly assigning subjects to the groups and as long as she has a lot of subjects, she has a good chance of coming out with equivalent groups.
But what if she didn’t have a lot of subjects? What if she only had 20 subjects, and could only put five people in each group? Then she’d want to do a multiple-subjects design with correlated groups, which is when Shakira uses non-random assignment to make her groups. The most common types of non-random assignment are matched groups, where Shakira will take all the shopaholics and distribute them evenly across the groups, and repeated measures, where Shakira will have every subject watch every commercial.
Every variable has levels, or options, for the value of that variable. If the independent variable of a study has only two levels, it is called a two-group design.
However, if the independent variable has more than two levels, it is called a multiple-group design. A two-group design answers whether a variable has an effect, while a multiple-group design answers how much of an effect each level of the variable has. Finally, multiple-group designs come in two forms: with independent groups, which is when subjects are randomly assigned to a group, and with correlated groups, which is when subjects are not randomly assigned to a group.
After watching this lesson, you should be able to:
- Define experimental design
- Understand what levels of a variable are
- Differentiate between two-group and multiple-group design
- Recall how independent and correlated groups are different