Validity: External, internal, and construct - Research Paper Example

? Validity: External, internal, and construct Introduction The establishment of authenti originality and applicability of any research work is a matter of judgment. The judgment is derived from the understanding drawn from various factors. There are mainly 2 different types of factors which are validity and reliability. Both the factors complement each other and both are indispensable in the successful establishment of the research efforts. Validity is related to mainly three different types of reasoning, which are external validity, internal validity and construct validity (Green, 2008). In order to establish the importance and difference between the three types of validity, a causal relationship between, loan default and poor credit worthiness is used as example. The research paper endeavors to uphold the contrasting features of external, internal and construct validity through subsequent comparison of the characteristics as well as threats for the same. Characteristics of external validity External validity examines if the causal relationship which is observed can be generalized and applied across different measures, persons, settings and times. This is one of the unique characteristics of external validity. If the conclusion derived from the test cannot be generalized then it means that it is applicable to only a part of the population (Linn, 2011). It is a type of evaluation of the experiment that questions whether the experimental results are applicable to the populations and situations which are different from the original experiment. For example, in the loan defaulter classification analysis it is found that certain characteristic features are unique to the defaulters. These characteristics can be applied to generate a probable score of the defaulters of short term loan (Lune, Parke & Stone, 2012). External validity confirms whether the test results can be applied to derive the credit scores of those defaulters who have taken long term loans. Characteristics of internal validity Unlike other validity tests like the construct validity and external validity, the internal validity indicates if the covariance observed between the variables can be at all regarded as true causal relationship. Internal validity helps to check if the experiments are performed correctly. It also helps to check if the confounds are occurring. If the internal validity of the experiment is found to be low then it indicates that there is high chance of presence of external variability. In the loan defaulter classification example a high covariance is found in between relaxation of the loan policy and the chances of loan default. It was found that during periods when the financial policies were relaxed higher numbers of loan defaults are witnessed (Moss, 2013). So it can be said that there is a direct covariance in the relaxation of the terms and policies and the frequency of loan default. Further investigation in the policies and terms of lending indicate that when the obligations to submit security deposits are relaxed then more number of defaulters is noticed. This evidently points to the fact that there is a definite cause and effect relationship between policy implementation and loan defaults (Reckase, 2010). Characteristics of construct validity Unlike other types of validity namely external validity and internal validity, construct validity usually was divided into 3 main types, which are content validity, criterion related validity and predictive validity. For example, in the debtors classification test based on the credit score one of the first criteria is the establishment of the fact that the sample used is a representation of the population (Taleporos, 2006). The criterion related validity helps to focus on the correlation of test with some standard measures. For example, the validity of the variables chosen to construct the model is compared with some other types of standardized measures. The variables chosen to built a similar score model by different bank or financial institution can be studies and then analyze the success rate of the model to test the predicting ability of the original model (Yen, 2005). The test can be further validated if the same statistical model is used to classify the defaulters and non defaulters utilizing a different set of data and then comparing the results of the actual experiment. Threats to external validity The threats to external validity can be divided into mainly two types, one is based on the methods used and the other one is based on participants. Based on methods used various kind of threats can be identified and one of them is threats from interaction of testing and treatment. Pre tests do not occur in the real world and thus the results may fail to generalize well. For example, banks have a credit score of the loan takers, although in most of the cases the credit scores are general in nature (Stigler, 2011). Other types of methods based threats are interaction of selection and treatment, reactive arrangements and presence of multiple treatment interfaces. In the case of participant based threats, the threats are limited type of population test, limited types of population tested, presence of gender bias, presence of racial bias and presence of cultural bias. For example, while extending credit terms to the debtors. In interaction of the selection and treatment, it may happen that the sample selected turn out to be homogenous in character, thus depleting the chances of applying the same to the general. Threats to construct validity Unlike the threats to the external validity, the threat to the construct validity can be divided into mainly two different types, which are Inadequate Preoperational Explication and the other one is the social threats. Due to preoperational explication, sometimes mono-operation bias may creep in. For example, in the selection of variables only those variables are considered which have fulfilled certain criteria. This may automatically make it imperative to use a single type of statistical test. This in turn can lead to mono-method bias. Mono-method bias may sometime creep in when only one method is applied to derive the results and conclusions instead of multiple methods. Due to restrictive generalizability the credit score derived from the process can term the rather credit worthy debtors as probable defaulters. This leads to what is known as unintended consequences. Another problem is the choice of confounding constructs and levels of construct. For example, the people who are marked as loan defaulters can be regarded as non defaulters with the use of different loan size (Green, 2008). The social threats to construct validity are hypothesis guessing, apprehension of the evaluation and expectancies of the experimenter. Due to wrong choice of the hypothesis the experiment can go haywire. For example, if the debtors are informed beforehand, that they are about to being included as part of a survey and experiment, they may become conscious and start exhibiting characteristics which are unusual. Impact of validity on envisioned research The internal validity which helps to establish the causal relationship between the reasons for default and probability of default is of prime importance. If the internal validity is not able to establish the causal relationship between the reasons for default and probability for default then it means that the factors identified as reasons do not have a direct relationship with the probability of default (Linn, 2011). In other words the experiment cannot pinpoint the probability of default in particular to any factor at all. Construct validity helps to establish if the factors and variables chosen for constructing the model can be linked theoretically. For example, it can be established theoretically in the credit score development project that the credit score of the individuals are related to the short term credit worthiness as well as long term credit worthiness. Though it cannot be established theoretically the reasons for which the debtors could have used the short term borrowing to finance the long term obligations (Lune, Parke & Stone, 2012). This kind of short coming is present and does not interfere with the actual result of the experiment. If the same shortcomings are observable among the main variables then the construct validation cannot be established. References Green, D. R., (2008). Consequential aspects of the validity of achievement tests: A publisher's point of view. Educational Measurement 27.2, pp.16-19.  Linn, R. L., (2011). Partitioning responsibility for the evaluation of the consequences of assessment programs. Educational Measurement 19.1, 28-30  Lune, S., Parke, C. S., & Stone, C. A., (2012). A framework for evaluating the consequences of assessment programs. Educational Measurement 14.6, 24-28.  Moss, P. A., (2013). The role of consequences in validity theory. Educational Measurement 17.22, 6-12.  Reckase, M. D., 2010. Consequential validity from the test developer's perspective. Educational Measurement 15.3, 13-16.  Taleporos, E., (2006). Consequential validity: A practitioner's perspective. Educational Measurement 18.9, 20-23.  Yen, W. M., (2005). Investigating the consequential aspects of validity: Who is responsible and what should they do? Educational Measurement 12.2, 5-6. Stigler, G. (2011). The theory of economic regulation. The Bell Journal of Economics and Management Science 2.7, 3–21. Read More