 
Abstract/Syllabus:

140.658
Statistics for Psychosocial Research: Structural Models
Description
Presents quantitative approaches to theory construction in the context of multiple response variables, with models for both continuous and categorical data. Topics include the statistical basis for causal inference; principles of path analysis; linear structural equation analysis incorporating measurement models; latent class regression; and analysis of panel data with observed and latent variable models. Draws examples from the social sciences, including the status attainment approach to intergenerational mobility, behavior genetics models of disease and environment, consumer satisfaction, functional impairment and disability, and quality of life.
Learning Objectives
Upon successful completion of this course, students will be able to design path analysis models; to analyze latent variable longitudinal data with linear structural equation models; to design latent class analysis models in the situation of categorical data; and to read and evaluate scientific articles as regards testing of causal relationships in public health based on a priori theory.
OCW offers a snapshot of the content used in courses offered by JHSPH. OCW materials are not for credit towards any degrees or certificates offered by the Johns Hopkins Bloomberg School of Public Health.
Syllabus
This syllabus corresponds to the Fall 2007 offering of Statistics for Psychosocial Research: Structural Models. It is not necessarily representative of subsequent offerings of the course.
Course Description
Presents quantitative approaches to theory construction in the context of multiple response variables, with models for both continuous and categorical data. Topics include the statistical basis for causal inference; principles of path analysis; linear structural equation analysis incorporating measurement models; latent class regression; and analysis of panel data with observed and latent variable models. Draws examples from the social sciences, including the status attainment approach to intergenerational mobility, behavior genetics models of disease and environment, consumer satisfaction, functional impairment and disability, and quality of life.
Course Objectives
Upon successful completion of this course, students will be able to design path analysis models; to analyze latent variable longitudinal data with linear structural equation models; to design latent class analysis models in the situation of categorical data; and to read and evaluate scientific articles as regards testing of causal relationships in public health based on a priori theory.
Prerequisites
MH 330.657 or equivalent. Auditing MH 330.657 is a sufficient entry criterion provided that the auditing student has completed the problem sets in that course. The course is designed to build upon what is learned in MH 330.657.
This course is the second in a twoquarter series on Statistics for Psychosocial Research. The series is oriented towards latent variable models and related methods and is taught jointly by the Departments of Mental Health and Biostatistics. The first quarter concentrates on measurement and the second quarter on structural models. The first quarter course, or permission of the instructor, is required for enrollment in the second quarter course.
Readings
Required:
Loehlin, JC. Latent Variable Models: An Introduction to Factor, Path, and Structural Analysis. Fourth edition. Hillsdale, NJ: Laurence Erlbaum Associates, 2004.
Highly recommended:
Maruyama, G. Basics of Structural Equation Modeling, SAGE, 1997.
Bollen, KA. Structural Equations with Latent Variables, New York: Wiley and Sons, 1989.
Course Requirements
Completion of three problems sets (each of which contribute 20% towards the final grade) and one inclass closebook final exam (which contributes 40% toward the final grade).
Attendance of weekly laboratory sessions is strongly recommended.
Schedule

1 
Introduction: Structural regression 
Model specification
Motivating examples
Three approaches: score then analyze, analyze then summarize, LV
Role of measurement error
Model assumptions
Path diagram


2 
Regression analysis for items 
Generalized estimating equations (GEE)/marginal models
Model specification, interpretation, and fitting


3 
Introduction to path analysis 
Path diagram
Decomposing covariances and correlations
Direct, Indirect, and Total Effects
Identification
Estimation


4 
Introduction to structural equations with latent variables 
Measurement models
Structural models
Model specification, Estimation
Example: confirmatory factor analysis


5 
Inference using structural equations with latent variables 
Parameterizing hypotheses
Parameter constraints
Model identification
Model checking


6 
Examples of path analysis 
Behavior genetics
Status attainment
Evaluation of treatment effects


7 
Commonly applied structural models with latent variables 
MIMIC (multiple indicators and multiple causes of a single latent variable) models
Group comparisons
Application (example)


8 
Advanced structural equations models I 
Longitudinal analysis
Growth curves


9 
Advanced structural equations models II 
Multilevel Models 

10 
Models for dichotomous outcomes 
Dichotomous variable factor analysis
Latent variable structural equations models with discrete data


11 
Latent class regression I 
Motivating examples
Model specification
Assumptions
Fitting


12 
Latent class regression II 
Model selection
Violations of assumptions
Identifiability
Model checking
Example

13 
Concluding topics 
Design, power, sample size
Pros and cons of latent variable models
Using observed and latent variable models in parallel
Causal inference




Further Reading:

Readings
This reading list corresponds to the Fall 2007 offering of Statistics for Psychosocial Research: Structural Models. It is not necessarily representative of subsequent offerings of the course.
Textbooks
Required:
Loehlin, JC. Latent Variable Models: An Introduction to Factor, Path, and Structural Analysis. Fourth edition. Hillsdale, NJ: Laurence Erlbaum Associates, 2004.
Highly recommended:
Maruyama, G. Basics of Structural Equation Modeling, SAGE, 1997.
Bollen, KA. Structural Equations with Latent Variables, New York: Wiley and Sons, 1989.

Introduction: Structural regression 
Reading
Loehlin, Chapter 1 "Path Models in Factor, Path, and Structural Analysis"
Additional Reading
Bollen, Chapter 1, "Introduction" and Chapter 2, "Model Notation, Covariances, and Path Analysis," pp. 1031
Bollen, Chapter 5 "The consequences of measurement error," pp. 151168

Regression analysis for items 
Optional readings
Diggle PJ, Heagerty P, Liang KY, Zeger SL. "Generalized Linear Models for Longitudinal Data" in Analysis of Longitudinal Data, 2nd edition, London: Oxford, 2002 (pp 5470). 
Introduction to path analysis 
Reading
Loehlin, Chapter 2 "Fitting path models"
Additional Reading
Bollen, Chapter 2 "Model Notation, Covariances, and Path Analysis," pp. 3239.

Introduction to structural equations with latent variables 
Reading
Loehlin, Chapter 3 "Fitting path and structural models to data from a single group on a single occasion," pp. 8795.
Additional Reading
Bollen, Chapter 7, "Confirmatory factor analysis," pp. 226238.

Inference using structural equations with latent variables 
Bollen, Chapter 7, "Confirmatory factor analysis," pp. 238296. 
Examples of path analysis 
Reading
Loehlin, Chapter 3, "Fitting path and structural models to data from a single group on a single occasion," pp. 95115
Additional Reading
Maruyama, Chapter 6, "Recursive and Longitudinal Models: Where Causality Goes in More Than One Direction and Where Data Are Collected Over Time," pg. 99108
Optional Reading
Sewell WH, Hauser RM, Wolf WC. "Sex, Schooling, and Occupational Status," American Journal of Sociology 1980;86:551583.
Wheaton B. "The Sociogenesis of Psychological Disorder: Reexamining the Causal Issues with Longitudinal Data," American Sociological Review 1979;43:383403.

Commonly applied structural models with latent variables 
Reading
Loehlin, Chapter 3, "Fitting path and structural models to data from a single group on a single occasion" pp. 98102.
Loehlin, Chapter 4, "Fitting models involving repeated measures or multiple groups," pp. 129148.
Additional Reading
Bollen, Chapter 8, "The general model, part I: latent variable and measurement models combined," pp. 355369.

Advanced structural equations models I 
Reading
Loehlin, Chapter 4, "Fitting models involving repeated measures or multiple groups," pp. 120129.
Additional Reading
Maruyama, Chapter 6, "Recursive and Longitudinal Models: Where Causality Goes in More Than One Direction and Where Data Are Collected Over Timeâ€?, pp. 108122.
Optional readings
Aneshensel CS, Frerichs RR, "Stress, Support, and Depression: A Longitudinal Causal Model," Journal of Community Psychology 1982;10:36376.
Muthen BO, "Analysis of Longitudinal Data using Latent Variable Models with Varying Parameters," in Best Methods for Analyzing Change

Advanced structural equations models II 
Suggested readings
Goldstein H, McDonald RP, "A general model for the analysis multilevel data," Psychometrika 1988;53:455467
Muthen BO. "Means and covariance structure analysis of hierarchical data," UCLA Statistics series, no 62: Los Angeles, 1990.
Muthen BO. "Multilevel factor analysis of class and student achievement components," Journal of Educational Measurement 1991;28:338354.

Models for dichotomous outcomes 
Reading
Bollen, Chapter 9, "The general model, part II: extensions," pp. 433447.
Additional readings
Muthen BO, "A general structural equation model with dichotomous, ordered categorical, and continuous latent variable indicators," Psychometrika 1984;49:115132.
Muthen BO, "Goodness of fit with categorical and nonnormal variables," in: Bollen KA, Long JS, eds., Testing structural equation models (pp. 205234). Newbury Park, CA: Sage.

Latent class regression I 
Additional readings
BandeenRoche K, Huang GH, Munoz B, Rubin GS, "On Determining Risk Factor Associations with Questionnaire Outcomes: A Methods Case Study," American Journal of Epidemiology, 1999;150:116578.
BandeenRoche K, Miglioretti DJ, Zeger SL, Rathouz PJ, "Latent Variable Regression for Multiple Discrete Outcomes," Journal of the American Statistical Association 1997;92:13751386.

Latent class regression II 
No reading 
Concluding topics 
Reading
Loehlin, Chapter 7, "Issues in the application of latent variable," pp. 98102.
Additional readings
Maruyama, Chapter 1, "What Does It Mean to Model Hypothesized Causal Processes With Nonexperimental Data?"
Bollen, Chapter 3, "Causality and causal models," pp. 4079.




Rating:
0 user(s) have rated this courseware
Views:
6303




