Types Of Validity In Research: Predictive Validity

We know that psychology uses the application of tests to make inferences about people. But, what do the scores obtained in these tests refer to? Validity takes care of this. In addition, as we have already explained in another article, there is not only one type of validity, but several. In this space we explain what predictive validity is responsible for .

Researcher Robert Thorndike (1989) explains that the validity of a test is related to what this test tries to measure. Thus, derived from this premise, it is worth asking: is the use or interpretation of the scores of this test adequate? Or also: What generalizations can be made without introducing a very large interval of error from the results obtained in the proof? Validity alludes to all these issues.

Definition of validity

In statistical terms, validity is defined as the proportion of the true variance that is relevant for the purposes of the test. What does the relevant term mean? In this case, we refer to the characteristics or dimensions that the test we are using is capable of measuring.

As we have mentioned previously, the validity of a test is defined:

• Through the relationship between their scores with some measure of external criteria. That is, if we have another test that measures the same, we would compare the results obtained with both, understanding that the new test measures the same as the first when the results obtained are similar.

• By the extent to which the test measures a hypothetical specific underlying trait or “construct”. In this case, a common practice is to compare the results of two parts of the test designed to measure the same thing. It is considered that the validity the higher the correlation / association between the results, the greater the validity.

Types of validity

As we have already mentioned previously, there are several types of validity. Specifically, the types of validity are:

• Content validity (or evidence of content).
• Predictive validity (or external evidence).
• Concurrent.
• Construct validity.

Predictive validity (or external evidence)

Do test scores predict future performance or behavior? (1) External validity takes care of answering this question.

Commonly in psychology tests are used to predict possible future behaviors. Thus, we use the test to help us make a practical decision (classification, selection, etc.). In each of these situations,  the higher the accuracy of the prediction, the higher the predictive validity of the test and, therefore, the more useful the test will be.  (1)

For example, let’s say we use a test in a recruitment case. In this case, the test will be an acceptable component of the personnel selection process as long as its scores predict the performance of some important component of the job to which the interviewees aspire. This would be called external criteria.

Thus, the researcher Jaime Aliaga says that for the test to be used as part of a personnel selection process, it is necessary, as you can imagine, that the test has a good validity. In this sense, the idea is to relate the test to the relevant criteria. So, it seems that the fundamental interest of the psychologist or evaluator is to determine whether the test being measured predicts a certain criterion.

To achieve this, says researcher Aliaga, we need the external criteria to which the test scores will be related to be reliable and valid criteria. But what are we talking about when we talk about criteria?

The criterion in predictive validity

According to Aliaga, a criterion is any performance that the subjects have in real life. For example, a criterion could be the measure of job performance or the measure of academic performance in the case of a student. We could talk about many more types of these criteria. The problem, says Aliaga, is that in many cases it is impossible to find an unambiguous criterion of a mental trait.

Let us say, for example, that two psychologists together investigate the academic performance of a certain number of students. Thus, investigating the same, the two psychologists can use different criteria. For example, the first psychologist might consider test scores as a criterion. The second could consider that the correct criterion is the time it takes for each student to complete each task.

When we want to validate a test, the predictive validity of a test (as well as the concurrent validity) is generally expressed by a correlation coefficient between the scores that we call criteria. This coefficient is called the validation coefficient. According to Aliaga, the interpretation of this coefficient requires an excellent command of the statistical analysis used to obtain this criterion. After obtaining the criterion, the next thing that will be important to obtain the validity category is the statistical procedures that are used.

What is predictive validity for?

Thus, predictive validity is one of the types we need to determine the validity of a test. Logically, and specifically in the field of psychology, a test is better the more valid it has. In addition, we must bear in mind that the key aspect of this type of validity is the criterion. We need tests that serve as reliable and valid criteria. With this, we can correctly determine the predictive validity.