Introduction
We may lack a universally accepted definition of validity but the concept that comes out from most of the definitions is the degree to which a concept under study is justifiable. According to the American Educational Research Association Psychological Association and National Council on Measurement in Education (1999), measurement validity is the extent to which the evidence available is sufficient to support the final reading of the test scores.
Levels of measurement
The measurement of data revolves around four main types of data (Nachmias & Chava 2007). Nominal data is where we have numbers only representing their respective labels. For instance, we might have a set of data in which we have five students, three professors and ten parents. The second type is the ordinal data. This type of data allows for ranking or determining of intensity. An example is a list of one’s priorities or a student’s score board. Thirdly, we have interval data. Ideally, it has continuums that have equally spaced intervals. This data also lacks a true zero point. An example of this would be the scores of an IQ test on a 5-point likert scale. Last but not least, we have ratios. This data has continuums that include absolute zero points. An example is the number of questions a student answers correctly in an exam.
The levels of measurement of data have little to do with establishing whether the data is continuous or discreet (Creswell, 2003). Levels of data measurement will dictate the statistics that can be computed and the statistical procedures that can be employed on the data. The process of transforming data ranges from taking physical logs to allow for normal distribution to recording of the data in order to pave way for easier data entry and analysis (Lewin & Somekh 2005). A practical example is a weather man on a news channel who might report, “last week it was an average of 35 but this week it’s an average of 70- twice as hot” In the example, the data has been transformed by multiplying the former value by two. This multiplication by two jeopardizes the relationship between 35 and 70.
Validity and reliability
What makes validity of data hard to establish is the fact that it is often limited to the accuracy and reliability of the data. Ideally, if the measurements presented are reliable, the data will probably be valid (Creswell, 2003). If the measurements are however not reliable, then validity of the data can not suffice. As Miller (2009) explains, if a measurement has high reliability, then the observed scores will be very close to their true scores. For one to be able to determine how reliable data is he/she will need to have a look at the variance of the scores in question. The variance from the true scores plus the variance from the measurement error will give us the total variance in the set of scores.
Conclusion
There is nothing half as important as realizing the measurement level of the data when one wants to make a right decision in statistical science. The individual usually holds the information concerning the relationship of the data and the property that he/she is interested in. In as much as statistical software like SPSS will readily compute a standard deviation on the ordinal data, the conclusions that will be arrived at after looking at the numbers might be bogus and this could ultimately result in making of wrong decisions.
References
American Educational Research Association, Psychological Association, & National Council on Measurement in Education. (1999). Standards for Educational and Psychological Testing. Washington, DC: American Educational Research Association.
Creswell, J. (2003). Research Design: Qualitative, Quantitative, and Mixed Methods Approaches. London: Sage Publications.
Lewin, C., & Somekh, B., (2005). Research methods in the social sciences. London: Sage Publications.
Nachmias, D., & Chava, F. (2007). Research methods in the social sciences. New York: Worth Publishers.