<abstract xmlns="http://www.w3.org/1999/xhtml"><p>Psychological constructs such as personality dimensions or cognitive traits are typically unobserved and are therefore measured by observing so-called indicators of the latent construct (e.g., responses to questionnaire items or observed behavior). The Common Factor Model (CFM) models the relations between the observed indicators and the latent variable. In this article we argue in favor of interpreting the CFM as a causal model rather than merely a statistical model, in which common factors are only descriptions of the indicators. When there is sufficient reason to hypothesize that the underlying causal structure of the data is a common cause structure, a causal interpretation of the CFM has several benefits over a merely statistical interpretation of the model. We argue that (1) a causal interpretation conforms with most research questions in which the goal is to <italic>explain</italic> the correlations between indicators rather than merely summarizing them; (2) a causal interpretation of the factor model legitimizes the focus on <italic>shared</italic>, rather than unique variance of the indicators; and (3) a causal interpretation of the factor model legitimizes the assumption of local independence.</p></abstract>

Psychological constructs such as personality dimensions or cognitive traits are typically unobserved and are therefore measured by observing so-called indicators of the latent construct (e.g., responses to questionnaire items or observed behavior). The Common Factor Model (CFM) models the relations between the observed indicators and the latent variable. In this article we argue in favor of interpreting the CFM as a causal model rather than merely a statistical model, in which common factors are only descriptions of the indicators. When there is sufficient reason to hypothesize that the underlying causal structure of the data is a common cause structure, a causal interpretation of the CFM has several benefits over a merely statistical interpretation of the model. We argue that (1) a causal interpretation conforms with most research questions in which the goal is to explain the correlations between indicators rather than merely summarizing them; (2) a causal interpretation of the factor model legitimizes the focus on shared, rather than unique variance of the indicators; and (3) a causal interpretation of the factor model legitimizes the assumption of local independence.

Toward a Causal Interpretation of the Common Factor Model

Disputatio

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

{"article-title":"Toward a Causal Interpretation of the Common Factor Model"}

Psychological constructs such as personality dimensions or cognitive traits are typically unobserved and are therefore measured by observing so-called...

Toward a Causal Interpretation of the Common Factor Model

Publié en ligne: 16 oct. 2018

Pages: 581 - 601

Reçu: 05 sept. 2017

Accepté: 02 nov. 2017

DOI: https://doi.org/10.1515/disp-2017-0019

Mots clés
Causality, reflective model, statistical model

© 2017 Riet Van Bork et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Toward a Causal Interpretation of the Common Factor Model

Publié en ligne: 16 oct. 2018

Pages: 581 - 601

Reçu: 05 sept. 2017

Accepté: 02 nov. 2017

DOI: https://doi.org/10.1515/disp-2017-0019

Mots clésCausality, reflective model, statistical model

© 2017 Riet Van Bork et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Mots clés
Causality, reflective model, statistical model