University of Warwick, Department of Sociology, 2012/13 SO 201

University of Warwick, Department of Sociology, 2012/13
SO 201: SSAASS (Surveys and Statistics) (Richard Lampard)
A glimpse beyond the module…
(Week 20)
Multiple Correspondence Analysis
• This isn’t exactly a straightforward extension of basic
correspondence analysis, but it does allow information
from multiple cross-tabulations to be pulled together to
‘unpack’ a mapping common to these. (See Chapter 7 of
Cox, T.F. 2005. An Introduction to Multivariate Analysis.
London: Hodder Arnold.)
• It has been used extensively by the French social theorist
Pierre Bourdieu, notably in his 1984 book Distinction.
• See also Harrits, G.S. 2013. ‘Class, culture and politics: on
the relevance of a Bourdieusian concept of class in
political sociology’, Sociological Review 61.1: 172-202.
Multi-level models
• These models are used when there are units
of analysis at two levels, e.g. pupils within
• However, sometimes the ‘higher’ level unit of
analysis is a ‘nuisance’ level, e.g. where a
sample of addresses is drawn from a sample
of areas
• MLMs are not a feature of SPSS version 21.
• Researchers use more flexible and/or
specialised software such as MLwiN.
Khattab, N., Johnston, R., Sirkeci, I. and Modood, T. 2012.
‘Returns on education amongst men in England and Wales: The
impact of residential segregation and ethno-religious
background’, Research in Social Stratification and Mobility 30.3:
Multinomial logistic regression
• In many ways this is a straightforward
extension of logistic regression.
• The dependent variable has three or more
categories, with one of these acting as the
reference category.
• An alternative is to break the analysis down
into binary logistic regressions.
Software again!
• Multinomial logistic regressions can be fitted
using SPSS (see Chapter 15 of Gray, C.D. and
Kinnear, P. 2012. IBM SPSS Statistics 19 Made
Simple. New York: Psychology Press.)