Experimental and Nonexperimental Designs Experimental

Experimental and Nonexperimental Designs

Experimental vs. non-experimental designs: An overview

An experiment involves the use of the scientific method: a researcher does preliminary research, constructs a hypothesis based upon the past knowledge he or she has surveyed, and then constructs a test or a study to prove or disprove the hypothesis. An experiment usually examines the effects of an independent variable upon a dependent variable. The independent variable is the variable that is controlled or manipulated by the researcher, and the dependent variable is the response or result (IV & DV, 2008, UNCP).

An example of an experiment might be the decision of a psychological researcher to study the effects of loud noise on students' ability to study. Some individuals claim that noise helps them study while others say they require silence. In the experiment, one group of students (evenly balanced between males and females and strong, average, and weak students) might take a test after studying in a loud room. The other group of demographically similar students would study for the test in a quiet room. The independent variable would be the noise; the dependant variable would be students' grades on the exam.

In contrast, a non-experimental design, such as a case study of students during their first year of college, would not isolate a particular influence upon the student's study habits and would examine the student's world from a phenomenological point-of-view. The researcher would observe the students, not manipulate variables. A variety of factors would be observed together, and the researcher would conduct interviews of the subjects, rather than focus on statistically verifiable data.

The existence of a control group in an experiment is designed to act as a means of comparison. The other experimental group or groups are compared with the control group, to demonstrate that there is a relationship between the independent and dependent variable. A control group can be used in experiments with more than one independent variable, although not every experiment uses control groups. For example, the researchers studying the effects of noise upon study habits might also wish to examine if all noise was treated the same by the brain when studying: one control group of students could study in silence, the other group would study in a room exposed to loud noise, while a third group of students could study in a room with quiet, ambient music. Or, the researchers could study musical preference and see if students who enjoyed rock n' roll were equally as affected by loud noise as students who preferred jazz, classical music, or folk music. This comparative study would have no formal control.

Scientists often use statistical analysis, called regression analysis, to plot the relationship between the different variables studied in experiments. For example, noise at a moderate level might enhance or not inhibit studying, while very loud, ear-splitting levels of noise might make study impossible (Rumsey 2007, p.80). A researcher could plot what noise level was optimal to achieve the best results when studying. Multiple regression analysis might be appropriate in some studies with multiple variables (Rumsey, 2007, p.88). To make the non-experimental anthropological study of freshman in a dorm room experimental, a study could be constructed of multiple variable factors (such as high school grades, number of roommates, major, gender) to determine what variables were present in students with higher grades. By submitting questionnaire to students in the dorm room and determining which variables, when present, seemed to correlate with higher grades and were potential 'causal' factors in higher grades, the researchers could then call their endeavor an experimental study. (Note that in this experiment there would be no formal control group, instead seeing what clusters of variables were present in high-achieving students would be the focus).

In general, the great advantage of experiments over other study designs is that, with an experimental design, you can more easily demonstrate cause-and-effect relationships. Doing a simple observational case study of students on a dorm hall would expose the researcher to many extraneous variables that could affect students' study habits all at once, and would make it difficult to compare his or her study with other surveys of factors that enhance student performance on tests.

However, constructing studies that meet the criteria for experiments is not always feasible. Sometimes, ethically, a researcher may wish to study the effects of certain variables, like that of long-term illegal drug use, which cannot support an experimental design. Following the lives and surveying vital physical data about volunteers who were drug users might be one way to accumulate important data. "Perhaps the simplest [non-experimental] design is the correlational design or quasi-experimental design. A study qualifies as correlational if the data lend themselves only to interpretations about the degree to which certain things tend to co-occur or are related to each other" (Non-experimental design, 2009, UNE).