Quasi-experiments are research designs that attempt to mimic an experimental design. However, the researcher cannot control the assignment of study participants to groups: a key feature for true experiments (Wadsworth, n.d.). This paper explores quasi-experiments as a research design.
Part 1
Advantages of Quasi-experiments
Firstly, quasi-experiments, as the name suggests, are not true experiments. They contain elements of both a true experimental and correlational research (Jackson, 2012). Therefore, they allow the researcher to obtain stronger conclusion compared to when they conduct a pure correlational research. Secondly, quasi-experiments are useful in circumstances when conducting a true experiment is either infeasible or very expensive. Thirdly, quasi-experiments are also useful in circumstances where ethical concerns prevent conducting experiments. Fourthly, quasi-experiments are conducted in a more natural setting compared to true experiments. Therefore, they have a high external validity.
Disadvantage
Firstly, although quasi-experiments allows the researcher to draw stronger conclusions than they would if the research design was correlational, the conclusion is not as strong as that of a true experimental design. Therefore, the findings of quasi-experiments are somewhat inconclusive as to the causal relationship (Jackson, 2012). There may be other explanations for the findings due to the existence of cofounds. Secondly, quasi-experiments have a low internal validity because the design includes a nonequivalent control group or a single group in some instances. Therefore, the possibility of alternative explanations cannot be ruled out.
The weakness of quasi-experimental designs matter. However, there are various mitigations that can be employed to minimize or eliminate the internal validity threats.
Random Assignment
In true experiments, the researcher assigns subjects randomly to the control and treatment group. The assumption of equivalence at the beginning holds for true experiments because the researcher is in charge of manipulating the treatment variable of interest. In true experiments, there is no equivalence at the end because manipulation has taken place. It is the differences between the control group and a treatment group that allows the research to confirm or disapprove the hypothesis of the causal relationship that is being studied (Jackson, 2012). In quasi-experiment, the researcher looks for naturally occurring treatment and control groups in the population. Therefore, the assumption of equivalence does not hold from the beginning. Therefore, random assignment is not possible. If a researcher is interested in analyzing the relationship between smoking and lung cancer they can look for smokers and non-smokers. Smoking cannot be manipulated by asking participants to smoke. The participants are recruited as either smokers or non-smokers.
The outcome of a study can suggest whether a particular threat exists. For instance, a study that reveals that lack of exercise causes obesity suffers from internal validity. From experience, there are various factors that cause obesity such as the quantity of food consumption, the nature of food and the timing of the food consumption.
Non-equivalent group design Post-test Only
It entails using naturally occurring intact groups in the population that the researcher assumes will reflect a control and treatment group after the treatment has already occurred (Trochim & Donnelly, 2008). For instance, an educational study will select two classrooms that are comparable in different schools to assess the effect on a certain teaching method on student learning. The researcher selects groups that share close similarities in terms of characteristics to minimize alternative explanations. The design solves some of the selection threats. However, it does not solve others. It solves selection-maturation because it measures the effect at a given point in time after treatment has occurred. It also solves for selection mortality because it is not conducted over an extended period. However, it does not solve for selection history. The differences in the groups from other factors are not eliminated. Selection instrumentation and regression threats will also persist.
Non-equivalent group design Pre-test/Post-test
This design observes two groups before treatment manipulation takes place and after the treatment manipulation takes place (JHA, 2014). For instance, a researcher maybe interested in assessing the effectiveness of a newly introduced teaching methods in certain schools. In this case, the researcher will select a class in a school where the teaching program has been introduced and select another one where the program has not been introduced. The researcher then compares performance before (pre-test) and after sometime when the program has been implemented (post) say one year. The pre-test/post-test design suffers from certain selection threats. There is selection history as there may be other factors that influence performance such as school resources, income or parents education level. There is selection-maturation because the student may have become better than before as they grow older. Selection mortality is also rife because it takes place over an extended period. Lastly, selection instrumentation is also likely to be a threat.
Cross-sectional design
A cross-sectional design is one that takes place at a given point in time (Rubin, 2012). It is the opposite of longitudinal which takes place over an extended period. A cross-sectional design solves the threats of selection maturity and selection mortality. Selection maturity takes places because the subject change with time for extended studies. However, the cross-sectional design takes place at a given time thus eliminates changes due to time. Selection mortality is a threat if the attrition rate between the two groups is different. Since cross-sectional studies take place in a single session, attrition is unlikely. However, cross-sectional design only solves for threats that are a function of time. Therefore, it is not a solution for selection-history, selection-instrumentation, selection-testing and selection-regression.
Regression Discontinuity
In the design, participants are assigned to groups by cut-off point basing a pre-study measure. For instance, studying the impact of a teaching method to students’ performance by selecting participants with a certain SAT score benchmark. The design solves partially for selection history. This is because the measure accounts for some of the study cofounds.
Reasons why Quasi-Experiments are Preferred to True Experiments
True experiments provided more conclusive results compared to quasi-experiments. However, quasi-experiments are preferred for various practical reasons especially in social sciences.
Firstly, there various instances when randomization is not possible because of ethical concerns (Babbie, 2007). For instance, in studying of the effect of alcoholism on aggressive behavior, a randomization would require that the treatment group is given excessive alcohol. It would be unethical. However, it alcoholics in the population can be identified and used as a control group.
Secondly, it may be too expensive to conduct experiments. A study on the effect of winning a lottery on happiness would require that the treatment group is given lottery with huge sums. This will be too costly. However, it is easier to locate lottery winners in the population.
Thirdly, quasi-experiments have a high external validity because the study is conducted in the natural setting. Therefore, the findings can be argued to be generalizable to the wider population.
Research Question
Does anger management classes influence aggressive behavior?
Rationale for the Study
The researcher was motivated by the perverseness of commercials to children that attracted extensive policy debates. In most countries, commercials that target children are subject to a strict regulatory framework. At the time, television commercial targeting children were banned in Quebec. Companies spend a lot of funds to advertise to children to enhance their sales. For instance, in 1988, at least 500 million dollars spent on advertisements targeting children (Goldberg, 1990). Therefore, the researcher sought to evaluate the effectiveness of TV commercials on children to assess whether it is justified.
The study was not designed to a contribution to theory because it does not develop any unique theory. Rather, it is applied research.
Variables
The independent variables are cultural affiliation that is proxied by language spoken and exposure to television commercials which are measured in hours. The dependent variable for the study is the response to commercials. The response is measured using awareness of toys (Goldberg, 1990).
Research Design
The researcher uses a quasi-experimental research design. The internal validity threat that exists is selection history bias arising from cultural differences. The researcher solves it by comparing English speaking and French speaking children who watch the same amount of American television. There is an external validity threat of generalization. The researcher solves it by including French speaking and English speaking children from different income groups to ensure a representative sample.
Findings
The researcher reveals that exposure to American television has a significant positive influence on the ability of children to recognize toys. The findings are convincing. There were two groups that were clearly separated by the variable of interest. The researcher solves the various validity threats that exist. Although the magnitude of the correlation may not be accurate, the nature of correlation is likely to be correct.
References
Babbie, E. (2007). The Basics of Social Research. London: Cengage Learning.
Goldberg, M. (1990). A quasi-experiment assessing the effectiveness of TV advertising directed to children. Journal of Marketing Research, 445-454.
Jackson, S. (2012). Research Methods and Statistics: A Critical Thinking Approach. London: Cengage Learning.
JHA. (2014). Social Research Methods. New Delhi: McGraw Hill Education (India) Pvt Ltd.
Rubin, A. (2012). Statistics for Evidence-Based Practice and Evaluation. London: Cengage Learning.
Trochim, W., & Donnelly, J. (2008). The Research Methods Knowledge Base. London: Cengage Learning.
Wadsworth. (n.d.). Quasi-Experiments. Retrieved from http://www.wadsworth.com: http://www.wadsworth.com/psychology_d/templates/student_resources/workshops/res_methd/non_exper/non_exper_03.html
