STATISTICS

Can an unreliable test be valid?

Can an unreliable test be valid?

Testing Reliability

Validity, reliability, and relevance are statistical concepts which are applied to measuring standards (Feldman, 2010). For example, a statistic is reliable if the results obtained can be duplicated conducting similar research. However, a statistic can be considered invalid (lacking validity) if the research did not measure what it was supposed to measure.

Reliability deals with accuracy and precision of the measurement procedure and validity is explained using two major forms of validity, which are internal and external. Internal validity is described as the ability of the research instrument to measure what is designed to measure and external validity is he data’s ability to generalized across persons, settings, and times (Feldman, 2010). For example, a measure of learning difficulties may be reliable but invalid if referring to only reading disabilities.

Reliability is a necessary, but not the only condition for establishing validity. For instance, a test may provide consistent answers meaning that it is reliable. However, the test would only be valid if the test was actually measuring what it was supposed to be measuring. For instance, a measure of intelligence using the size of brains might provide consistent results but this would not make it valid.

This brings into play another idea which is the difference between precision and accuracy. It is entirely possible for data to be precise but inaccurate. Accuracy is more important because is the “closeness between measurements (observations) and their expectations (“true” values)” (Stanford, 2015). The data can be said to be inaccurate as it moves away from the expected value. If the data is precise then it means that it is close with regard to how it is observed. The closer data is when observed or calculated than the greater the precision. However, if data is close in observation or calculation this does not mean that it is accurate. This works both ways but arriving at data which dictates inaccuracy is not helpful due to the fact that it is not showing an expected value.

Within this framework relevance becomes important to data. Data can be used to show correlations with disabilities but only if the data is relevant. For example, if the number of reading disabilities are diminishing and learning disabilities are increasing, this does not mean that there is a correlation in the data. This is due to the fact that these disabilities may not be related with one another.

References

Feldman, R. S. (2010). Psychology and your life. New York, NY: McGraw Hill

Photo by Chris Liverani on Unsplash

~Citation~

Triola Vincent. Tue, Feb 09, 2021. Can an unreliable test be valid? Retrieved from https://vincenttriola.com/blogs/ten-years-of-academic-writing/can-an-unreliable-test-be-valid

Need similar articles?

Statistics
Back to: Ten Years of Academic Writing