Data Collection Strategies
34 Nonexperimental Research
Nonexperimental research is research that lacks manipulation of an independent variable and/or random assignment of participants to conditions. While the distinction between experimental and nonexperimental is considered important, it does not mean that nonexperimental research is less important or inferior to experimental research (Price, Jhangiani & Chiang, 2015).
When to use nonexperimental research
Often it is not possible, feasible, and/or ethical to manipulate the independent variable, nor to randomly assign participants to conditions or to orders of conditions. In such cases, nonexperimental research is more appropriate and often necessitated. Price, et al. (2015) provide some examples that demonstrate when the research question is better answered with non-experimental methods, as follows:
- The research question or hypothesis contains a single variable rather than a statistical relationship between two variables (e.g., How accurate are people’s first impressions?).
- The research question involves a noncausal statistical relationship between variables (e.g., Is there a correlation between verbal intelligence and mathematical intelligence?).
- he research question involves a causal relationship, but the independent variable cannot be manipulated, or participants cannot be randomly assigned to conditions or orders of conditions (e.g., Does damage to a person’s hippocampus impair the formation of long-term memory traces?).
- The research question is broad and exploratory, or explores a particular experience (e.g., What is it like to be a working mother diagnosed with depression?).
As demonstrated, above, it is the nature of the research question that guides the choice between experimental and non-experimental approaches. However, this is not to suggest that a research project cannot contain elements of both an experiment and a non-experiment. For example, nonexperimental studies that establish a relationship between two variables can then be further explored in an experimental study to confirm or refute the causal nature of the relationship (Price, Jhangiani & Chiang, 2015).
Types of nonexperimental research
In social sciences it is often the case that a true experimental approach is inappropriate and often unethical. For example, conducting a true experiment may require the researcher to deny needed treatment to a patient, which is clearly an ethical issue. Furthermore, it might not be equitable or ethical to provide a large financial or other reward to members of an experimental group, as can occur in a true experiment.
There are three types of nonexperimental research: single-variable research; correlational and quasi-experimental research; and, qualitative research. In the following sections we will explore single variable research and correlational research and quasi-experiments. Qualitative research is a large body of research methods that will be more fully discussed in the following chapters.
Single-variable research
Single-variable research is, not surprisingly, focused on a single variable, as opposed to a statistical relationship between two variables (Price et al., 2015). As Price et al. (2015) point out, Stanley Milgram’s original obedience study (See Chapter II, Research on Human Subjects: An Historical Look) is an interesting example of research focused on one variable—the extent to which the participants obeyed the researcher when he told them to provide a shock to the person behind the screen. Milgram observed all participants undertaking the same task, under the same conditions.
Correlational research
Correlational research is a type of nonexperimental research in which the researcher is interested in the relationship between variables; however, the researchers does not attempt to influence the variables (in contrast to experimental research where the researcher manipulates the variables) (Siegle, 2015). Relationships between variables can be visualized with the aid of a graph known as a scatterplot diagram.
Scatterplots provide information on two dimensions. The first dimension demonstrates the direction of relationship: linear, curvilinear, or no relationship. Linear relationships can be positive or negative. A positive relationship or correlation is demonstrated through a rise from left to right, while a negative correlation falls from left to right (Palys & Atchison, 2014). Here is a short video that effectively demonstrates positive relationships and no relationship: Example of Direction in Scatterplots.
The second dimension related to scatterplots is that they can provide an indication of the magnitude or strength of the relationship. The strongest of relationships are evidenced when all points in a scatterplot graph fall along the same straight line (known as the regression line). The next strongest of relationships are evidenced by a little bit of dispersion around the line; however, if one were to draw an oval close to the line all points would be captured within the oval. The more dispersed the points (i.e. the points do not closely adhere as closely to the line), the weaker the relationship (Palys & Atchison, 2014).
Near the beginning of the 20th century, Karl Pearson developed a method to statistically measure the strength of relationships between variables. This method, known as the Pearson product-moment correlation coefficient (Pearson’s r), was developed to measure the strength of linear relationships only. There are two aspects to Pearson’s r : The first is the direction, represented by a sign (+ or −). A plus sign (+) indicates a positive or a directional relationship, while a negative sign (−) indicates a negative or an inverse relationship. The second aspect is a number, where a zero represents no linear relationship, and a 1.0 represents a perfect linear relationship. A 1.0 is represented on a scatterplot when ever point lies on the same straight line. For these purposes, we will not delve further into how to compute a correlational coefficient; however, there are many online and library statistical resources if you wish to seek more information on this measure.
Quasi-experiments
Under certain conditions, researchers often turn to field experiments, also known as quasi-experiments (Crump, Price, Jhangiani, Chiang, & Leighton, 2017). Quasi-experiments allow researchers to infer causality by using the logic behind the experiment in a different way. There are three criteria that must be satisfied for causality to be inferred:
-
- the independent variable (X) comes before the dependent variable (Y) in time;
- X and Y are related to each other (i.e., they occur together);
- the relationship between X and Y aren’t explained by other causal agents.