Item Response Theory (TRI)

Assessment is one of the most important parts of psychological intervention. This is often heavily conditioned by the results of the tests administered. Thus, item response theory (IRT) is one of the test measurement theories that appear as a complement to classical test theory.

As we mentioned before, classical test theory (TCT) and TRI could evaluate the same test. Each could establish a relevance or score for each of the items. Thus, it would be possible to obtain a different result for each person. It is noteworthy that TRI would give us a much better calibrated instrument. However, this paradigm is associated with a much higher cost and the participation of specialized professionals.

The goal of these two test theories is the same: to generate instruments that measure what we want them to measure with as little error as possible. Thus, psychometry requires a certain reliability and validity for all tests.

Remember that a test will be more reliable (it will have greater reliability) the better it replicates the results when measuring two subjects – or the same subject at different times – that have the same level of measurement. On the other hand, validity refers to the degree to which empirical evidence and theory support the interpretation of test scores.

Limitations of TCT that led to the emergence of Item Response Theory

Without neglecting its service, the classical test theory approach has some limitations. These are shortcomings that require us to advance in terms of building and evaluating the tests.

In TCT, measurements are not invariant to the instrument used. So imagine that a psychologist will assess the intelligence of three people with a different test for each. In this case, the results could not be compared. Why?

This is because each test has its scale. Thus, in order to be able to compare, for example, the intelligence of a group of people who were assessed with different intelligence tests, it would be necessary to transform the scores obtained on other scales.

In this sense, the TRI allows us to compare the results obtained when using different instruments on the same scale. Furthermore, another limitation of the classical approach is the lack of invariance of test properties in relation to the people used to calculate them. The TRI approach is also responsible for improving this fact.

Assumptions of Item Response Theory (TRI)

In order to resolve these limitations, the TRI has to make stronger and more restrictive assumptions than the TCT.

first guess

The most important assumption of the IRT is that any measurement instrument must be aligned with an idea, that there is a functional relationship between the values ​​of the variable that measure the items and the probability of combining them. This function is called the item characteristic curve (CCI).

It appears that item-response theory proposes a new idea about TCT. This is based on the fact that, for example, the most complicated items on an intelligence test would be answered only by those who are more intelligent. On the other hand, an item answered in the same way by all the people evaluated would not have the power to discriminate between more or less intelligent in a subject.

second guess

Another IRT assumption is that most models assume that items constitute a single dimension. That is, they are one-dimensional. So, before using the models of this theory, we must make sure that the data conforms to this one-dimensionality. This poses an important restriction on their use: many of the instruments that psychologists manipulate do not collect single-dimensional data.

third guess

A third assumption of item-response theory models is local independence. This means that to use these templates, items must be independent of each other. That is, the answer to one of them cannot be conditioned to the answer given to other items.

However, if unidimensionality is met, local independence is also met (there is no interdependence of items or a shared variance that is not related to the measured dimension). So sometimes both assumptions are handled together.

Muñiz (2010) points out the importance of advances in the field of psychometrics and test interpretation. So, the logical thing is that we start taking another step in this direction, since the tests analyzed under the IRT paradigm show, at least, worrying results about how the measurement is currently being done.