0
$\begingroup$

Explanatory item response models add predictors of the $\theta$ "ability" parameter. However, these predictors are always observed covariates. Sometimes, one may want to include latent variables as predictors, such as a different dimension representing a different type of ability. This is the item response theory equivalent to a structural equation model.

What is the name for this model? Any resources suggested for learning more?

$\endgroup$
1
  • $\begingroup$ in educational assessment context this is often done using plausible value representations of said constructs. $\endgroup$
    – Tom
    Oct 1, 2022 at 11:16

2 Answers 2

1
$\begingroup$

See Boeck and Wilson (2004): https://link.springer.com/book/10.1007/978-1-4757-3990-9

There is no special name for this. It is just an (possibly double) explanatory item response model.

$\endgroup$
1
$\begingroup$

To clarify, you are asking for information regarding an IRT model where a latent variable $\theta_y$ can be regressed onto one or more latent predictor variables $\theta_x$, correct? I am unaware of any formal presentation of such a model, though it should be pretty straightforward in a Bayesian framework using probabilistic programming languages (e.g., jags, stan).

One article (Lu, Thomas, & Zumbo, 2005) does come close to what you are looking for. First, they estimate the SEM's measurement model parameters using IRT. Second, they convert IRT parameters to their CFA/SEM parameterization$^1$; and third, they keep these parameters fixed during the estimation of structural parameters using SEM software.

$^1$ See Takane & De Leeuw (1987) for more information on the relationship between IRT and factor analysis of categorical data.

References

Lu, I. R., Thomas, D. R., & Zumbo, B. D. (2005). Embedding IRT in structural equation models: A comparison with regression based on IRT scores. Structural Equation Modeling, 12(2), 263-277.

Takane, Y., & De Leeuw, J. (1987). On the relationship between item response theory and factor analysis of discretized variables. Psychometrika, 52(3), 393-408.

$\endgroup$
3
  • $\begingroup$ Come across the first paper quite a bit but nowhere subscribes to that journal! $\endgroup$ Oct 1, 2022 at 1:12
  • $\begingroup$ Note: this appears do-able within the context of latent-variable linear mixed models. See page 383 here routledge.com/Longitudinal-Data-Analysis/… $\endgroup$ Oct 1, 2022 at 2:44
  • $\begingroup$ Glad it helped! I don’t have access to the book you referenced, but I will take a look when I get the chance. $\endgroup$ Oct 3, 2022 at 16:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.