The Cathie Marsh Centre for Census and Survey Research

Latent Factor Analysis

Book on Introduction to Data Analysis Part 1 Dates: 16th November 2011
Duration: 1 day (10am — 4:30pm)
Level: Intermediate
Course Fee: £175 (£125 for those from educational institutions)

CCSR offers 5 free places to research staff and students within the Faculty of Humanities at the University of Manchester and the North West Doctoral Training Centre.

Course Leader: Wendy Olsen
Course Requirements: A basic knowledge of syntax (commands) in a statistical package such as SPSS, and a previous exposure to regression analysis, are both required. 

The course aims to present the results of factor analysis using SPSS and MPLUS software, but some STATA graphics may also be used.  Participants can, if they wish, use this course as part of a move from basic SPSS use toward more advanced use of ‘syntax’ in statistical work.  However the syntax is provided for each exercise, making the course easier to follow.

Course Summary

This short course covers latent variables and factor analysis at an introductory and intermediate level.  A latent variable is a thing (such as an attitude, an orientation, an experience or a level, e.g. the level of well-being) that has been measured using a set of related indicators.  A set of three or more indicators can be considered the manifest variables, from which a single latent variable might be derived.  Factor analysis is one way to derive a single factor from a set of variables, and is thus called a data reduction method.  Other data reduction methods include principal components analysis, which is very closely related to factor analysis, and multiple correspondence analysis.  We will focus on confirmatory factor analysis, but during the course we also define and explain exploratory factor analysis. The course is suitable both for primary-data collection researchers (who may need to write a suitable questionnaire), and for those who want to analyse secondary data sets.

Course Aims

The short course is tightly focused.  It aims:

  • to show what kinds of models would lead to an adequate confirmatory factor model
  • to distinguish a one- from a two-factor model (or more factors) 
  • to help participants become familiar with two or three exemplars from social science research where latent variables are useful 
  • to see how non-continuous variables (those measured by nominal or ordinal, ranked measurement) can be put into new software (MPLUS) to generate a continuous latent factor
  • to introduce the concept of a test of goodness of fit.

 (The examples offered are likely to include attitude scales, a human capabilities index or an index of well-being, and models of the division of labour by gender.)

Note:  The course does not cover latent class analysis.  In latent class analysis, the results would give discrete classes which optimally separate the cases into groups according to the values of the manifest variables.  Once MPLUS is used, it is relatively easy to move from latent factor analysis to latent class analysis.  Both techniques can then be used in more advanced contexts.

Software Used 

The course is not strongly dependent on any one computer package, but participants are shown the SPSS and MPLUS methods of running confirmatory factor analysis. 

Content of Workshop

Morning: Introductory Lecture:  Theory of Confirmatory Factor Analysis (CFA).  Controversial Aspects. Practical Session in SPSS.  Attitude Scale Vs. Latent Factor of Attitudinal Orientation.  Example of Attitudes.  Tests of Fit.

Afternoon:  Conclusion about Traditional CFA.  Visualising a Factor Analysis.  Requirements for CFA to Work with Continuous Variables. Practical Session.  An Index of Human Capabilities (UK). Lecture on New Forms of CFA.  Practical Session.  Running and Viewing MPLUS Output.

Preliminary Reading

  • Hair, J.F., Anderson, R.E., Tatham, R.L. and Black, W.C. (2005). Multivariate Data Analysis. New Jersey: Prentice-Hall.
  • Loehlin, J.C. (2004). Latent Variable Models: An Introduction to Factor, Path, and Stuctural Equation Analysis, 4th ed. NY: Psychology Press.
  • Tabachnick, B.G. and Fidell, L.S. (1996). Using Multivariate Statistics. New York, NY: HarperCollins College Publishers. (see also 5th edition 2006)

Exemplar

  • Fuller, B., Caspary, G., Kagan, S.L., Gauthier, C., huang, D.S.C., Carroll, J. and McCarthy, J. (2002). 'Does maternal employment influence poor children's social development?' Early Childhood Research Quarterly 17: 470-497

 

 

Book on Introduction to Data Analysis Part 1

University of Manchester CCSR