3.3: Research Design

Generally quantitative studies examine the relation between two or more variables. Variables are characteristics of the individual or situation that can vary in amount or type. For example age, weight, self esteem and ethnicity are all individual variables. Type of child care situation, public or private funding for schools are situational variables. Generally, for any hypothesis, the variable of interest that is assumed to be the cause is called the independent variable in experimental design studies. The outcome variable is called the dependent variable in experimental design studies. In strictest terms the terms “independent and dependent variables” should only be used in experimental designs, whereas you can use the terms “predictor and outcome variables” to refer to the same variables in quasi-experimental settings. Quantitative research studies are different depending on the amount of control that the researcher can exert on the variables in question. At one extreme is experimental design –

Experimental Design

When you have a hypothesis about the effects of one variable on another, in experimental designs, the researcher manipulates or changes one variable in order to observe the effects on the other variable. For example, if a researcher wanted to study the effects of phonics teaching versus whole language teaching on reading outcomes for K-3 students, in experimental research, the researcher would randomly assign some students to the phonics condition and some students to the whole language condition, and then test their reading levels after having been exposed to the different conditions for some specified period of time.

The steps might look something like this

1. Find 200 Kindergarten students
2. Randomly assign 100 to learn using phonics, and 100 to learn using whole language
3. Two months later administer the same reading test to all 200 children

Quasi Experimental Design

However, if the researcher merely went to two schools where each of the methods – phonics versus whole language - was employed already in each of the schools, and tested children’s reading levels, it might appear that this would be a simpler way to get the same (or similar) results. But you can never be sure. In fact, in the second situation, the researcher has no control over other confounding variables that might systematically vary along with one or both of the variables of interest. For example, it might be that the school where teachers use whole language to teach reading is located in a higher SES area than the one that employs phonics methods. So, when we test children’s reading at both schools, we can’t be sure whether reading levels differ due to the method of teaching or the SES. This kind of study would be considered quasi experimental.

It might also be difficult for the researcher to truly assign individuals to different IV conditions when the hypothesized cause is ethnicity or sex for example. A researcher cannot randomly assign participants to male and female groups or to Black and White groups. Instead they would try to match the two groups in all other ways – e.g. age, SES, and other things that might influence the outcome. The more variables that are controlled or accounted for by the researcher, the more confidence we can have that any relation found between the two variables of interest is truly a valid one. This is also important because there are some variables that cannot be controlled because of ethical considerations.

Can you think of variables that it would be unethical for a researcher to manipulate? Can you think of how those variables do in fact vary and affect other developmental factors? How might researchers study these kinds of conditions?

Correlational Design

Correlational design is another type of research design at the other extreme from experimental designs. Here the researcher only measures two variables. There is no manipulation of any of the variables. For example, a researcher might wish to measure whether children’s aggression levels vary with local temperature, or whether children’s self esteem is related to use of strict discipline. The steps followed in a correlational design are as follows

1. Identify a school where you want to conduct the study
2. Collect data about
   1. Temperature each day
   2. How many kids were sent to the office for aggressive conduct with classmates each day
3. Run statistical analysis to see if there are patterns

In this case, you might find that as temperature increases, the numbers of children sent to the office for hitting their classmates increases. This is considered a positive correlation between the two variables. In the second example above, you might find that the stricter parents report being, the lower their children’s self esteem. This would be considered negative correlation between the two variables. The term correlation is used in slightly different ways in different contexts so it’s important to note those differences:

• Correlational design is a type of research design where the researcher exists minimal control and only measures two variables.
• Correlation coefficient refers to a statistical calculation that results in a number often denoted by “r.” The Pearson Product Moment Correlation coefficient is a number that ranges between 0 and the absolute value of 1 (-1--+1). The direction of the relation is indicated by the sign, and the strength of the relation by its closeness to 1. So a -0.9 correlation indicates a stronger relation between the two variables in question than a +0.3 correlation.
• Correlation refers to the relation between two variables and you will often hear it said that correlation is not causation. This follows from the above explanation of this type of design (ie correlational design versus experimental design). Since the researcher does not control any variables, given a correlation between two variables A and B, it is difficult to tell whether A caused B, B caused A, or some third variable C caused A and B to appear to be related.

Attributions:

Child Growth and Development by Jennifer Paris, Antoinette Ricardo, and Dawn Rymond, 2019, is licensed under CC BY 4.0

Research Methods in Developmental Psychologyby Angela Lukowski and Helen Milojevich is licensed under a CC BY-NC-SA 4.0