Next:
Preliminary info
Modelling Categorical Data: Loglinear models and logistic regression
Brendan Halpin,
G&S, Limerick University
August 26-30, 2002
Loglinear Analysis Preliminary info
Preliminary info
Overview
Models
Software
Modelling Categorical Data
Introduction: the analysis of categorical data
What's Categorical Data?
Poisson Regression
Binary dependent variables
Binary regression
Models for categorical data
Tabulation and association
Higher dimensions
Independence
Measures of association
The Odds ratio
Data representations
Loglinear models for 2-way tables
Analysing the structure of tables
Tangent: GLMs
Independence in loglinear terms
Loglinear allows us other models
What these models mean
Association
The
distribution
Probabilities are areas under the curve
Calculating probabilities
Fitting models in SPSS
Goodness of fit
Comparing models
Calculating model fit
Parameter estimates
Parameter estimates for higher dimensions
Interactions
Tables with more than 2 dimensions
The structure of 3-D tables
Interpreting 3-D models
Conditional vs Marginal Association: an example
Death penalty example
Odds Ratios and what they mean
Processes generating odds-ratios
Fitting models to large tables
Residuals
Loglinear models for square tables
Square Tables
Fitting models to square tables
Symmetry
Quasi-Symmetry
Weighting and cell-specific fitting
Other models
Fitting and interpreting models
Vote intentions
Models for ordered categories and trends (i)
The Linear-by-Linear model
Score models
Log-multiplicative models
More dimensions: trend models
The logit/loglinear equivalence
Logistic regression
Logistic through loglinear
Multinomial logit models through loglinear
Models for ordered categories (ii)
Using
EM
Linear and log-multiplicative models
Some special topics
Collapsibility
Collapsing categories
Sparseness and sampling zeros
Zero cells
Dependence and unobserved heterogeneity: overdispersion
Sigmoid curves and binomial distibutions
Grouped and individual level logistic
Multinomial and Ordinal Logistic Regression
Ordinal Dependent Variables
The Proportional Odds Model
© Brendan Halpin
(e-mail)
14-May-2003
Department of Sociology
,
University of Limerick
distribution
EM
