ISTQL 2012

ISTQL – 2012
Latent Class Analysis Spatial Econometrics
Allan L. Mccutcheon Anil Kumar Bera
Latent Class Analysis

This course provides participants with a practical, working knowledge of one of the major approaches to categorical data analysis and market segmentation research: latent class analysis. The latent class model (LCM) provides one of the most important approaches for understanding relationships among a wide variety of respondents and consumer attitudes and behaviours. The LCM facilitates the construction and testing of models of respondent types (e.g., market segments), and helps in characterizing the nature of complex groups. In addition to the basic LCM, attention is given to modelling longitudinal data, as well as to the inclusion of predictor and outcome variables related to the segments. We examine the conceptual and methodological foundations of market segmentation, with several specific applications of the latent class and related models to a variety of data types.

The course introduces participants to the conceptual and methodological foundations of the latent class model and its linkages to market segmentation, as well as its application to a variety of related social and political science problems. The expectation is that, at the end of the course, participants will be able successfully to conduct research using latent class analysis, as well as be prepared to pursue self-directed study in the area.

Course Outline
Latent Class Analysis

Allan L. McCutcheon

Week 1


Class: An Introduction to the Latent Class Model: Causal and Noncausal Analysis

[Reading: McCutcheon (1987a), Chapters 1 & 2]

-Continuous Data: Regression and Factor Analysis Models

-Categorical Data: Loglinear and Latent Class Models

-Latent vs. Manifest Variables

-Basic Concepts: conditional and Latent Class Probabilities; Local Independence

-Logic and notation of multi-way cross-classifications

-Basic Model Parameters: Specification and restrictions

Tutorial: Exploratory modelling of selected cross-clarifications


Class: Unrestricted and Restricted Models

[Reading: McCutcheon (1987a), pg. 27-35, 37-44, McCutcheon 2002]

-The Logic of Confirmatory Models

-Restricting Conditional and Latent Class Probabilities

Tutorial: Model restrictions with LEM


Class: Model Selection and Evaluation

[Reading: McCutcheon (1987a), 35-37]

-Partitioning Hierarchical Likelihood Ration Chi-squares

-Additional measures and cell assignment strategies

Tutorial: Model analysis and selection with LEM


Class: Scale Analysis: The Logic and an Example

[Reading: McCutcheon (1987a), Chapter 4]

-The logical foundations of scale analysis

-Restricted latent class models for scale analysis


Tutorial: LEM restrictions for scale analysis



Class: Simultaneous Latent Structure Analysis: The Logic and an Example

[Reading: McCutcheon (1987a), Chapter 5 & McCutcheon (1987b), Pg. 256-275]


-The logic of multiple group comparisons

-Across-group restrictions


Tutorial: Simultaneous latent structure models with LEM



Week 2



Class: Simultaneous Latent Structure Analysis: Model Selection

[Reading: McCutcheon and Hagenaars, 1997]


-Heterogeneity, structural homogeneity, and distributional homogeneity

-Trend analysis with simultaneous latent structure models


Tutorial: Simultaneous latent structure modelling of individually selected cross-classifications



Class: Loglinear Structural Equation Modelling: The Logic and an Example 

[Reading: Hagenaars (1993), 1-19, McCutcheon and Mills (1998)]


-The logic of causal models with latent variables

-Causal models as restricted latent class models


Tutorial: Specifying causal models as restricted latent class models with LEM



Class: Loglinear Structural Equation Modelling 

[Reading: Hagenaars (1993), 20-39]


-Causal systems with manifest and latent variables

-The latent class model as a restricted loglinear model


Tutorial: Specifying causal models with latent categorical variables using LEM



Class: Loglinear Structural Equation Modelling: Advanced Topics 

[Reading: Hagenaars (1993), 39-64, Hagenaars (2002)]


-Complex causal systems with manifest and latent variables


Tutorial: Complex causal models with latent categorical variables using LEM



Class: Advanced Topics:  Models for Analyzing Categorically-Scored Panel Data and

Hierarchical (Multi-level) Latent Class Models

[Reading: McCutcheon, (1997), Vermunt (2005)]

Tutorial: None


Participants should have some understanding of basic probability and contingency tables (crosstabulation), as well as basic algebra. A solid understanding of the logic of causality, such as found in regression and factor analysis, is also necessary. The mathematics/statistics underlying the logic of regression and factor analysis are useful, but not essential.

Davis, J. A. 1986. The Logic of Causal Order. Sage. QASS No.55.

McCutcheon, A., and Mills, C. 1998. ‘Categorical data analysis: Log-linear and latent class models’, in E. Scarbrough and E. Tanenbaum (Eds.), Research Strategies in the Social Sciences. Oxford University Press.
Spatial Econometrics
From the University of Illinois at Urbana-Champaign Anil K. Bera’s course is entitled “Spatial Econometrics”. The course will have two parts: Spatial Statistics and Spatial Econometrics. The aim of this course is to build up a strong theory foundation and to provide the computer application on spatial statistics and econometrics. The course starts from an introductory level and covers the frontier of research. It would be better if the participants in this course have strong Statistics & Econometrics background.

ISTQL 2012 Spatial Statistics and Econometrics Summer School Course Outline

Prof. Anil K. BERA

University of Illinois at Urbana-Champaign

Department of Economics

  1. Spatial Analysis
  1. The Beginning: Spatial Analysis in Time of Cholera
  2. Why (Spatial) Dependence Matter?
  3. Differences Between Time-Series and Spatial Dependence
  4. Measures of Spatial Dependence and Approaches to Spatial Modeling
  5. Simultaneous Autoregressive (SAR) and Conditional Autoregressive (CAR) Models
  6. Testing Spatial Dependence: Moran’s I and Other Tests
  7. Spatial Stationarity and Isotropy
  8. Variogram Analysis
  9. Spatial Prediction (Kriging)
  10. An Introduction to Social Interaction Models
  1. Spatial Econometrics
  1. Spatial Autoregressive and Error Models
  2. Estimation of Spatial Econometric Models (MLE, IV, GMM)
  3. Testing Spatial Models (with an Introduction to “Testing Statistical Hypothesis: LR, W and LM(Rs) Tests”)
  4. Emprical Applications
  1. Crime in Columbus
  2. Data with negative autocorrelation
  3. House prices in cities (Boston, Chicago, …)
  4. Reflection problem (social interaction and its effects ….)
  5. Data from China, Turkey, Europe
  6. Democracy around the world
  1. Spatial Panel Models
  1. Standard Panel Models
  2. Fixed and Random Effects
  3. Spatial Panel Models
  4. Estimation of Spatial Panel Models (ML, IV and GMM Methods)
  5. Testing Spatial Panel Models
  6. Empirical Applications

a.    Growth convergence (Turkey, Europe, etc.)
b.    International trade overtime.

Allan L. McCutcheon
Allan L. McCutcheon holds the Donald O. Clifton Chair of Survey Science, and is Professor of Statistics and Survey Research & Methodology at the University of Nebraska at Lincoln; he is also the founding Director of the Gallup Research Centre and founding Chair of the Doctoral program in Survey Research & Methodology. He has taught workshops on categorical data analysis and public opinion polling at the ICPSR Summer Program at the University of Michigan; the Central Archive for Empirical Social Research at the University of Cologne; the Central European University in Budapest and the Catholic University of Brussels. He is, with Jacques Hagenaars, co-editor of Applied Latent Class Analysis (Cambridge University Press. 2002), and author of Latent Class Analysis (Sage 1987). He has also published several articles applying log-linear and latent class models to substantive issues in social, political, and cross-cultural research.  He is an elected fellow of the American Statistical Association and the Royal Statistical Society.

Allan L. Mccutcheon

Anil K. Bera
Anil K. Bera was born in a remote village Paschimchak, West Bengal, India. He attended his village schools, Narendrapur Ramkrishna Mission College, and the Indian Statistical Institute, Calcutta and Delhi. Anil received his Ph.D. in Econometrics from the Australian National University. Before joining the University of Illinois in 1983, he was a CORE Fellow at the Université Catholique de Louvain, Belgium.

Anil is named to the List of Teachers Rated as Excellent almost every semester he teaches. He received the Economics Graduate Students. Organization (EGSO) Award for Excellence in Graduate Teaching eight times since 1989, the College of Commerce Alumni Association Outstanding Teaching Award for Graduate Teaching in 1991 and Honorable Mention of the Campus Award for Excellence in Graduate and Professional Teaching in 2005. Even after so many years of living outside, Anil is very nostalgic about his village. He visits there regularly, and is currently engaged in some development projects, such as building a Free Library and a Primary School building.

Anil Kumar Bera