Introduction
The concept of validity was first originated by Stanley and Campbell in 1966 and later elaborated further by Campbell and Cook in 1979. The concept of validity has a significant impact on how researchers evaluate their work. Validity refers to the extent to which to which a given test measures that which it claims to measure. It is important to evaluate the validity of a test to ensure accurate interpretation and application of test results. There are four main types of validity; construct validity, statistical validity, external validity and internal validity. This paper seeks to compare and contrast the characteristics of external, internal and construct validity. It further discusses the threats to external and constructs validity in light of my envisioned research. My envisioned research seeks to examine the United States educational system and more importantly focus on the role of for-profit colleges in today’s education system in increasing success rate for students.
Internal Validity
Internal validity measures the extent to which there are causal relationships between two variables, the explanatory variable (X) and the dependent variable (Y), irrespective of what the two variables represent theoretically. Internal validity is, therefore, appropriate for experimental studies that seek to identify causal relationships between variables. Internal validity is not appropriate for descriptive or observational studies. In experimental studies, unlike descriptive or observational studies, the explanatory variable is under the researcher’s control and he/she can manipulate it in so as to observe its influence on the dependent variable. Unlike, external and construct validity, internal validity does not attempt to generalize. The central question that internal validity seeks to address is whether the changes observed in the dependent variables are attributable to the program intervention.
In my envisioned research, internal validity would be irrelevant. This is because given the nature of the research; a true experimental design is not possible. It is also not possible to do a regression analysis to identify causal relationship when assessing the role of for- profit colleges in the United States education system in increasing success rate for students.
Construct Validity
Construct validity measures the extent to which legitimate inferences can be made given the operationalization of a research to the theoretical framework on which the operationalization was based. Similar to external validity, construct validity relates to generalizing. However, external validity relates to generalizing from the study setting to other places, times or people whereas construct validity relates to generalizing from the study setting to the theoretical concepts of the study program or measures. Both internal validity and construct validity are appropriate for experimental studies and not descriptive or observational studies. However, internal validity is only concerned with the ascertaining the validity of causal relationships whereas construct validity is interested in the interpretation of the theoretical framework or construct behind the causal relationships.
There are several threats to construct validity. The first threat is the researcher’s bias. The researcher transfers his/her expectations to the research participants to an extent that it influences the performance and a variation of the dependent variables. The second threat is condition diffusion. There is a possibility that participants who are in different condition groups can communicate during evaluation. Their communication will influence the results that will be obtained from both groups. The third threat is resentful demoralization. One group that is not receiving anything discovers that a treatment which another group is getting is effective. The fourth threat is insufficient preoperational explication. During a research, the construct used does not accurately assess what the researcher wants and it incorporates other constructs that should be separated from the desirable construct. The last threat is mono-operation bias. It arises from using one dependent variable in assessing a construct which results in underrepresentation of the construct.
In my envisioned research, construct validity would be irrelevant. This is because given the nature of the research; an experimental design is not possible. It is also not possible to do a regression analysis to identify causal relationship between the variables of interest since the study will be descriptive. Consequently, how threats to construct validity would be overcome was not discussed.
External Validity
External validity is the estimated truth of conclusions that involve generalizations. External validity relates to generalizing from the study setting to other places, times or people. The central concern of external validity is representativeness and heterogeneity of the of the evaluation sample used by a researcher to the population. Both external validity and construct validity relate to generalizing.
There are several threats to external validity. The first threat is a small size. A small sample size may not be representative of the entire population. Therefore, there is a need for the researcher to use an appropriate sample size. In my envisioned research, I will pick an appropriate number of for-profit colleges in the U.S.A to constitute a sample which will be determined using the scientific method. The second threat is the interaction effect of testing. Pre-testing of a sample causes interaction between the sample and the experimental treatment which influences the results in a way that the obtained results cannot be generalized to a population that was not pre-tested. The third threat is the interaction effect of selection bias. The result obtained from a sample obtained from a given segment of a heterogeneous population cannot give a valid generalization of the entire population. In my envisioned research, I will overcome this threat by selecting a sample randomly. The fourth threat is the reactive effect of an experimental setting. This threat arises due to the fact that the study subject are cognisant that they are participating in an experiment; had they been in a natural setting and unaware they are being observed they could have behaved differently. This threat will not be applicable to my envisioned research because the research will not involve an experimental design.
In my envisioned research, external validity would be very relevant. Generalizing the role of for-profit colleges in the USA education system in increasing success rate for students from my sample would be important to me as a researcher. However, the study cannot be used to generalize the roles of colleges in the USA because the researcher is only interested in a given segment of colleges in the USA (for-profit colleges) which will constitute the sample.
Conclusion
There are four main types of validity; construct validity, statistical validity, external validity and internal validity. Internal validity measures the extent to which there are causal relationships between two variables. Construct validity measures the extent to which legitimate inferences can be made given the operationalization of a research to the theoretical framework. External validity is the estimated truth of conclusions that involve generalizations. Unlike internal validity, external and construct validity relate to generalizing. Both internal and construct validity relate to causal relationship. Given the nature of my envisioned research only external validity will be relevant.
References
Lissitz, R. W. (2009). The Concept of Validity: Revisions, New Directions and Applications (Hc). Chicago: IAP.
Babbie, E. R. (2010). The Practice of Social Research (12 ed.). New York: Cengage Learning.
Healey, J. F. (2011). Statistics: A Tool for Social Research (9 ed.). London: Cengage Learning.
Murphy, K. R. (2011). Validity Generalization: A Critical Review. New York: Taylor & Francis,.
Reis, H. T., & Judd, C. M. (2010). Handbook of Research Methods in Social and Personality Psychology (revised ed.). Cambridge: Cambridge University Press.
