Further Statistical Methods - Lecture 6 and 7
Latent variable models
The source for these lecture is the excellent little book, in particular chapters 1, 3 and 4.
D. J. Bartholomew and Knott (1999). Latent Variable Models and Factor Analysis. Griffin, London.
It is describing a number of different methods for describing complex dependencies among several variables through the variation in a few common latent (unobservable) factors. Data on the factors are then missing in a very strong sense, and certainly MCAR, so the EM algorithm becomes a useful tool.
The general setup is well described in Chapter 1, the normal linear factor model in Chapter 3, and the binary latent trait model in Chapter 4.
Overheads
Links
Last updated: Thursday, 15 February 2007 09:44. Steffen L. Lauritzen