Log-Linear Modelling of Contingency Tables
Date:
30/05/2013 - 31/05/2013
Organised by:
NCRM
Presenter:
Professor John 'Mac' McDonald
Level:
Entry (no or almost no prior knowledge)
Contact:
Jacqui Thorp
Training Administrator
National Centre for Research Methods
University of Southampton
Email: jmh6@soton.ac.uk
Map:
View in Google Maps (SO17 1BJ)
Venue:
University of Southampton
Building 39
Highfield
Southampton
Description:
PLEASE NOTE THIS EVENT WAS ORIGINALLY ADVERTISED AS 25-26 MARCH IN MANCHESTER. THIS HAS NOW BEEN CHANGED TO 30-31 MAY IN SOUTHAMPTON.
Data analysts are often interested in describing the association(s) between cross-tabulated categorical variables. Log-linear models are commonly used for modelling contingency tables, where the cell counts are treated as the response variable. A classic example is a 2-way social mobility table, where father's social class is cross-classified by son's social class, and we are interested in describing the association (interaction structure). Standard models for square tables discussed include quasi-independence, quasi-symmetry and symmetry as well as newer models which impose interpretable structures on the interactions, such as the UNIDIFF and row-column association models from sociology. For higher-dimensional tables, log-linear models can be used to study whether 2-way, 3-way or higher-way associations exist among the variables and to estimate their strength of association. Course emphasis will be placed on the usefulness and interpretation of the log-linear models presented using ideas such as conditional independence and mathematical graphs which visually represent the conditional independence structure between variables.
The course is suitable for those who are completely new to R, but some practice with R before the course is strongly recommended (a worksheet will be provided). Familiarity with linear regression will be assumed, but some basic knowledge of logistic regression and/or log-linear modelling is desirable as well as familiarity with the concepts of odds and odds ratios.
Cost:
The fee is:
1. £60 - For UK registered postgraduate students
2. £120 - For staff at UK academic institutions, ESRC funded researchers and registered charity organisations
3. £440 - For all other participants
All fees include event materials, lunch, morning and afternoon tea. They do not include travel and accommodation costs. Refunds for cancellation are as follows. Full refund for cancellation one month before the course, NO refunds can be made after this date.
Website and registration:
Region:
South East
Keywords:
Survey and Questionnaire Design, R software , Log-Linear regression
Related publications and presentations:
Survey and Questionnaire Design