A Framework of Statistical Tests For Comparing Mean and Covariance Structure Models

Although statistical procedures are well-known for comparing hierarchically related (nested) mean and covariance structure models, statistical tests for comparing non-hierarchically related (nonnested) models have proven more elusive. Although isolated attempts at statistical tests of non-hierarchically related models have been made, none exist within the commonly used maximum likelihood estimation framework, thereby compromising these methods' accessibility and general applicability. Building on general theory developed by Vuong (1989) and techniques for establishing the relation between covariance structure models (Raykov & Penev, 1999), this work provides a general paradigm for conducting statistical tests on competing mean and covariance structure models. The proposed framework is appropriate for hierarchically related models as well as non-hierarchically related models. In developing the structure of the framework, key aspects of model equivalence, relation, and comparison are unified. An illustration demonstrates its use.

[1]  N Cliff,et al.  Some Cautions Concerning The Application Of Causal Modeling Methods. , 1983, Multivariate behavioral research.

[2]  H. Akaike.  Factor analysis and AIC , 1987 .

[3]  H. Bozdogan.  Model selection and Akaike's Information Criterion (AIC): The general theory and its analytical extensions , 1987 .

[4]  G. Schwarz.  Estimating the Dimension of a Model , 1978 .

[5]  Jean-Francois Richard,et al.  The encompassing principle and its application to non-nested hypotheses , 1986 .

[6]  J. H. Steiger.  Statistically based tests for the number of common factors , 1980 .

[7]  R. Golden.  Discrepancy Risk Model Selection Test theory for comparing possibly misspecified or nonnested models , 2003 .

[8]  Roland T. Rust,et al.  Comparing covariance structure models: A general methodology , 1995 .

[9]  R. Bagozzi,et al.  Assessing method variance in multitrait-multimethod matrices: The case of self-reported affect and perceptions at work. , 1990 .

[10]  B. Byrne,et al.  Modeling MTMM Data from Additive and Multiplicative Covariance Structures: An Audit of Construct Validity Concordance. , 1993, Multivariate behavioral research.

[11]  Kevin A. Clarke.  Testing Nonnested Models of International Relations: Reevaluating Realism , 2001 .

[12]  R. P. McDonald,et al.  Goodness of Fit in Item Response Models. , 1995, Multivariate behavioral research.

[13]  Peter M. Bentler,et al.  EQS : structural equations program manual , 1989 .

[14]  Kenneth A. Bollen,et al.  OBSERVATIONAL RESIDUALS IN FACTOR ANALYSIS AND STRUCTURAL EQUATION MODELS , 1991 .

[15]  Tenko Raykov,et al.  Approximate Confidence Interval for Difference in Fit of Structural Equation Models , 2001 .

[16]  Q. Vuong.  Likelihood Ratio Tests for Model Selection and Non-Nested Hypotheses , 1989 .

[17]  R. Rust,et al.  Incorporating prior theory in covariance structure analysis: A bayesian approach , 1989 .

[18]  S. Hershberger,et al.  A Simple Rule for Generating Equivalent Models in Covariance Structure Modeling. , 1990, Multivariate behavioral research.

[19]  T Raykov,et al.  On Structural Equation Model Equivalence. , 1999, Multivariate behavioral research.

[20]  D. Cox.  Tests of Separate Families of Hypotheses , 1961 .

[21]  H. White.  Maximum Likelihood Estimation of Misspecified Models , 1982 .

[22]  I. Stelzl.  Changing a Causal Hypothesis without Changing the Fit: some Rules for Generating Equivalent Path Models. , 1986, Multivariate behavioral research.

[23]  Jean-Francois Richard,et al.  The Encompassing Principle and Its Application to Testing Non-nested Hypotheses , 1986 .

[24]  Golden,et al.  Statistical Tests for Comparing Possibly Misspecified and Nonnested Models. , 2000, Journal of mathematical psychology.

[25]  C. Gouriéroux,et al.  Chapter 44 Testing non-nested hypotheses , 1994 .

[26]  J. Pearl.  Causality: Models, Reasoning and Inference , 2000 .

[27]  Terry E. Duncan,et al.  Latent Variable Modeling of Longitudinal and Multilevel Substance Use Data. , 1997, Multivariate behavioral research.

[28]  Roderick P. McDonald,et al.  What can we learn from the path equations?: Identifiability, constraints, equivalence , 2002 .

[29]  David C. Schmittlein,et al.  A Bayesian Cross-Validated Likelihood Method for Comparing Alternative Specifications of Quantitative Models , 1985 .

[30]  George A. Marcoulides,et al.  Can There Be Infinitely Many Models Equivalent to a Given Covariance Structure Model? , 2001 .

[31]  K. Widaman.  Hierarchically Nested Covariance Structure Models for Multitrait-Multimethod Data , 1985 .

[32]  S. Reise,et al.  Assessing the fit of measurement models at the individual level: A comparison of item response theory and covariance structure approaches. , 1999 .