IMPS 2008: Keynote Lectures

Keynote Lectures

Thomas Griffiths (UC Berkeley)		The Modern Intuitive Statistician
		The idea that human inferences can be characterized by the principles of Bayesian statistics has been championed and derided at various points in the history of cognitive science. In this talk, I will argue that this idea is still valuable in making sense of one of the most impressive aspects of human reasoning: our ability to solve inductive problems, such as learning causal relationships or learning categories based on only a few observations. In these cases, Bayesian statistics gives us a tool for understanding why people reach the conclusions that they do, and for identifying the assumptions that guide their inferences. In particular, I will focus on how tools from modern Bayesian statistics, such as hierarchical Bayesian inference, nonparametric latent variable models, and Markov chain Monte Carlo, can be used to shed light on some of the sophisticated inferences that people make in everyday life.

David Kenny (Univ Connecticut)		Fixed and Random Effects Communicating to Each Other: Examples from Dyadic Research
		Most models have two pieces: a fixed and a random piece. Normally in modeling little or no attention is played to the coordination of these two parts of the model. I suggest how these two different types of effects can be related using examples from my research area of dyadic analysis. First, I discuss how specification error in one part of the model affects the other part of the model. As an example, I use the measurement of consensus in person perception. Second, I discuss how a fixed effect is often closely tied to given random effect and I argue that an analysis of the random effects can be used to re-conceptualize a fixed effect. As an example, I consider the effect of group diversity on group cohesiveness. Third, I consider the relative power of tests of fixed and random effects and I argue that fixed effects often have considerably more power than random effects. As an example, I consider the measurement of gender differences in nonverbal sensitivity.

Bengt Muthén (UCLA)		Exploratory Structural Equation Modeling
		Exploratory factor analysis (EFA) has been said to be the most frequently used multivariate analysis technique in statistics. In 1966 Jennrich solved a significant EFA factor loading matrix rotation problem by deriving the direct quartimin rotation. He was also the first to develop standard errors for rotated solutions although these have still not made their way into most statistical software programs. This is perhaps because Jennrich's achievements were partly overshadowed by the 1967 development of confirmatory factor analysis (CFA) by Joreskog. Joreskog developed CFA further into structural equation modeling (SEM) where CFA was used for the measurement part of the model. The strict requirement of zero cross-loadings in CFA, however, often does not fit the data well and has led to a tendency to rely on extensive model modification to find a well-fitting model. In such cases, searching for a well-fitting measurement model may be better carried out by EFA (Browne, 2001). Furthermore, misspecification of zero loadings tends to give distorted factors with over-estimated factor correlations and subsequent distorted structural relations. This paper describes an EFA-SEM (ESEM) approach, where in addition to or instead of a CFA measurement model, an EFA measurement model with rotations can be used in a structural equation model. The ESEM approach has recently been implemented in the Mplus program. ESEM gives access to all the usual SEM parameter and the loading rotation gives a transformation of structural coefficients as well. Standard errors and overall tests of model fit are derived. Geomin and target rotations are discussed. Examples of ESEM models include multiple-group EFA with measurement and structural invariance testing, test-retest (longitudinal) EFA, EFA with covariates and direct effects, and EFA with correlated residuals including EFA of traits in MTMM settings. Testing strategies with sequences of EFA and CFA models are discussed. Simulated and real data are used to illustrate the points.