Do our rulers still rule?
Symposium organised by Denny Borsboom, Department of Psychology, University of Amsterdam, The Netherlands
Chair: Denny Borsboom, Thursday 23rd July, 11.30 - 12.50, Palmeston Lecture Theatre, Fisher Building.
Kees-Jan Kan, Rogier Kievit, Conor Dolan, Han van der Maas, Department of Psychology, University of Amsterdam, The Netherlands. On the interpretation of the CHC Gc factor as Crystallized Intelligence.
Rogier Kievit, Jan-Willem Romeijn, Lourens Waldorp, H. Steven Scholte, Denny Borsboom, Department of Psychology, University of Amsterdam, The Netherlands. Causality, Structural Equation Modeling and Cognitive Neuroscience: Psychometric Modeling of Reductive Science. ♥
Denny Borsboom, Lourens Waldorp, Angélique Cramer, Han van der Maas, Conor Dolan, Department of Psychology, University of Amsterdam, The Netherlands. What if there were no latent variables? Complex systems and psychometric models.
Angélique Cramer, Lourens Waldorp, Han van der Maas, Denny Borsboom, Department of Psychology, University of Amsterdam, The Netherlands. Comorbidity: A network perspective.
ABSTRACTS
Session overview
Denny Borsboom
Psychometric models have seen significant developments in the past decades, and are becoming increasingly popular in many areas of scientific psychology. However, there are many unresolved issues in the foundations of such applications. Such issues relate to the questions of how 1) to interpret central concepts in psychometric theory (e.g., the notion of a ‘latent variable’) in substantive terms, and 2) whether the basic assumptions underlying latent variable models can actually be expected to hold in the typical research settings in which they are applied (i.e., research into intelligence, personality, and psychopathology). The presentations in this session address such questions with regard to the interpretation of crystallized intelligence (Kees-Jan Kan), the relation between neuroscientific measures and psychological test scores (Rogier Kievit), the use of latent variable models in the analysis of comorbidity among mental disorders (Angélique Cramer), and the general question of whether latent variable models apply to systems that involve reciprocal causal relations between variables of interest (Denny Borsboom).
On the interpretation of the CHC Gc factor as Crystallized Intelligence
Kees-Jan Kan, Rogier A. Kievit, Conor V. Dolan, Han L.J. van der Maas
In Cattell’s (investment) theory of fluid and crystallized intelligence, fluid intelligence is a unitary relation perceiving capacity, which is related to the maturation of the cortex. The theory states that during the course of development people usually ‘invest’ their fluid intelligence to acquire domain specific abilities (knowledge, skills, solving strategies), which are called crystallized abilities. Consequently, crystallized intelligence is defined as the sum of these particular crystallized abilities. The concepts of fluid and crystallized intelligence have been influential not only in developmental psychology, but also in hierarchical psychometric models of intelligence. Within these models, the higher order factors denoted as Gf and Gc are referred to as fluid and crystallized intelligence, respectively. Although investment theory can explain why some people acquire more crystallized abilities than others, it fails to explain why such abilities still show positive correlations after fluid intelligence (Gf) is taken into account, and thus give rise to a common factor (Gc) which is supposed to represent crystallized intelligence. If one interprets factors as representing sources of individual differences and if one accepts investment theory, then it follows that Gc cannot represent crystallized intelligence, but a second common influence over and above the influence of fluid intelligence. In line with Cattell (1971, 1987), we propose that Gc represents individual differences in ‘exposure to education’. Keywords: Crystallized intelligence; Investment theory; Gf-Gc theory; CHC model; Factor analysis; Latent variable modeling.
Causality, Structural Equation Modeling and Cognitive Neuroscience: Psychometric Modeling of Reductive Science
Rogier A. Kievit, Jan-Willem Romeijn, Lourens J. Waldorp, H. Steven Scholte, Denny Borsboom
Cognitive neuroscience commonly involves the simultaneous analysis of behavioral and neurological data. Most contemporary studies in cognitive neuroscience ignore the implicit causal relationship between the distinct classes of data, and use simple correlations that do not allow for substantive interpretation. We argue that theoretically grounded integration of psychological and physiological theory is necessary for conceptually grounded reductive neuroscience, and so far has been thoroughly neglected. We show how two philosophical theories that discuss the relationship between psychological/behavioral states and brain states, namely identity theory and supervenience theory, allow for analogous translation into psychometric models and as such are empirically testable. Such an approach is insightful as it allows for the establishment of the conceptual, statistical and empirical foundations of cognitive neuroscientific research. By fitting MIMIC and simple reflective models to various datasets we illustrate the possibilities for integrating two distinct classes of data in such a way that substantive causal interpretations of latent constructs are possible. Furthermore, we will explore the possibilities for extending and applying these and other models to a wide range of cognitive neuroscientific topics. Keywords: Structural equation modeling; Cognitive neuroscience; Latent Variable Modeling; Conceptual analysis.
What if there were no latent variables? Complex systems and psychometric models
Denny Borsboom, Lourens J. Waldorp, Angélique O.J. Cramer, Han L.J. van der Maas, Conor V. Dolan
In many situations where latent variable models are applied to model a set of observables, substantive considerations suggest that these observables do not depend on a common latent variable, but instead are likely to influence each other directly, often reciprocally. If this is the case, then such variables form a causal network, the temporal trajectory of which is described by dynamical systems theory. The question discussed in this presentation is how such causal networks relate to psychometric models. It is shown that, under certain conditions, complex systems yield data that are empirically indistinguishable from those generated by IRT models. In particular, an interesting equivalence exists between a restrictive subset of causal networks and the Rasch model. In less restrictive situations, the data generated by complex systems are likely to be almost, but not exactly, unidimensional. Simulations show that the departures from unidimensionality are, however, so small that they are unlikely to be picked up in regular IRT modeling. The question whether such systems do or do not instantiate a latent variable model is discussed. Keywords: Item Response Theory; Dynamical systems theory; Reciprocal causation; Causal networks.
Comorbidity: A network perspective
Angélique O. J. Cramer, Lourens J. Waldorp, Han L. J. van der Maas, Denny Borsboom
The pivotal problem of comorbidity research lies in the psychometric foundation it rests on, i.e., latent variable theory, in which a mental disorder is viewed as a latent variable that causes a constellation of symptoms. From this perspective, comorbidity is a (bi)directional relationship between multiple latent variables. We argue that such a latent variable perspective encounters serious problems in the study of comorbidity, and offer a radically different conceptualization in terms of a network approach, where comorbidity is hypothesized to arise from direct relations between symptoms of multiple disorders. We propose a method to visualize comorbidity networks and, based on an empirical network for major depression and generalized anxiety, argue that this approach generates realistic hypotheses about pathways to comorbidity, overlapping symptoms, and diagnostic boundaries, that are not naturally accommodated by latent variable models: some pathways to comorbidity through the symptom space are more likely than others, those pathways generally have the same direction (i.e., from symptoms of one disorder to symptoms of the other), overlapping symptoms play an important role in comorbidity, and boundaries between diagnostic categories are necessarily fuzzy. Keywords: Latent variable theory; Psychopathology; Comorbidity; Complex systems analysis. (181)