Measuring Variables

Measuring variables can be a complicated task, and selecting valid measures is a complex undertaking (Sechrest, 2005). Some levels of measurement can be transformed to other levels, although others cannot. Special problems are posed by applying measurements in the social sciences, especially as they play a significant role in the inferences we, as social scientists make. Sechrest (2005) suggests the validity of the data is more important than the validity of a specific instrument, although selecting a valid instrument contributes to valid data.

The four levels of measurement include nominal, ordinal, interval, and ratio (Fort-Nachmias & Nachmias, 2008). The level of measurement determines the analysis appropriate for the numbers. Higher level measurements can be measured at lower levels. For example, a ratio can also be transformed to a nominal measurement, although the opposite is not true; a nominal measurement cannot be transformed to an interval or a ratio measurement. Nominal data, such as male versus female cannot be transformed to an interval, ordinal, or ratio scale. There is not enough information in the measurement; people fit into one category or the other. Being male or female neither captures information on rank order, as in ordinal scales, nor does being in a male or female category indicate meaningful differences between the categories. Male or female categories do not convey information related to having a zero point, as in a ratio scale. Therefore, male or female data, which is nominal, cannot be transformed to any of the other three levels of measurement.

On the other hand, ages of people is ratio level data. It can be given a number, there is an order to age, there are meaningful difference between the data (a 75 year old man is 2.083 years older than a 36 year old woman), ratios can be calculated because an individual cannot be any younger than 0. Data can be transformed according to the amount of information captured within the measurement. Whereas nominal levels of measurement capture a limited amount of information, ratio level data conveys a higher complexity of information. The higher levels of measurement have the same properties as the lower levels of measurement and can be transformed to lower levels, although the lower levels of measurement do not have the same properties so cannot be transformed to higher levels of measurement.

Whereas the concept of design validity is the extent to which research is sound (Fort-Nachmias & Nachmias, 2008), measurement validity is the extent to which the instrument measures what it intends to measure. A valid research design utilizes methods that result in relatively true findings. A valid design relies on internal and external validity to give credence to its results and the inferences made from them. In studies that aim to establish causal relationships, internal validity is the strength of the inferences made regarding these relationships (Trochim, 2006). External validity is the study's generalizeability to wider populations (Fort-Nachmias & Nachmias, 2008; Trochim, 2006).

Within the concept of measurement validity, three basic types of validity exist: content validity is the extent to which the instrument utilized for measurement measures the salient characteristics of the subject being measured. Empirical validity determines the degree to which the results produced by the measurement tool are similar to real world relationships between the variables (Fort-Nachmias & Nachmias, 2008). The extent to which a measure "correlates with some other phenomenon in which one is interested" (para. 1) determines its empirical validity (Sechrest, 2005). Construct validity isthe extent to which an instrument measures what it claims it is measuring and not something else (Cronbach & Meehl, 1955; Trochim, 2006). It is necessary for the researcher to find information appropriate to each of the three types of validity to validate the measurement.

References

Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological Bulletin, 52(4), 281-302. doi: 10.1037/h0040957

Frankfort-Nachmias, C., & Nachmias, D. (2008). Research methods in the social sciences (7th ed.). New York: Worth.

Marques, S., & Lima, M. L. (2011). Living in grey areas: Industrial activity and psychological health. Journal of Environmental Psychology, 31(4), 314-322. doi:10.1016/j.jenvp.2010.12.002

Sechrest, L. (2005). Validity of Measures Is No Simple Matter. Health Services Research, 40(5p2), 1584-1604. doi: 10.1111/j.1475-6773.2005.00443.x

Trochim, W. M. (2006). Quasi-experimental design. The Research Methods Knowledge Base.
Retrieved December 21, 2013, from http://www.socialresearchmethods.net/kb/quasiexp.php