- Home
- GENERAL NON-FICTION REVIEWS
- Latent Variable Models & Factor Analysis - A Unified Approach Reviewed By Manoj Rengarajan of Bookpleasures.com
Latent Variable Models & Factor Analysis - A Unified Approach Reviewed By Manoj Rengarajan of Bookpleasures.com
- By Manoj Rengarajan
- Published February 12, 2012
- GENERAL NON-FICTION REVIEWS
Manoj Rengarajan
Reviewer Manoj
Rengarajan holds a Master of Financial Engineering - University of
California, Berkeley and he works in the investment management
industry and specializes in providing economic and investment outlook
and strategy for global equity and government bond markets. He has an
educational background in financial engineering, business, and
engineering, and professional interests include business,
finance, economics, technology and related areas.
Publisher: Wiley
Blackwell
Author: David J. Bartholomew, Martin Knott and
Irini Moustaki
ISBN: 978-0470971925
Latent Variable Models
and Factor analysis presents a comprehensive and unified approach
using a general framework to factor analysis and latent variable
modeling. While the book assumes no prior background in the topic,
the depth covered is impressive covering recent developments
including the use of MCMC methods and structural models.
Latent
variable models attempt to explain complex relations between several
variables by simple relations between the variables and an underlying
unobservable. The book presents a general framework to derive
commonly used models and provides several numeric examples to discuss
the practical applications of the techniques.
Starting with
the generalized linear latent variable other models are discussed as
extensions. Even the normal linear factor model is explained as a
special case of the general model. The fitting of the models by
maximum likelihood estimation and Bayesian methods are discussed. An
important aspect of the third edition is the introduction of MCMC
methods for parameter estimation.
Latent variable modeling for
binary and polytomous data is covered in depth including maximum
likelihood estimation of the model, fitting logit/ normal and probit/
normal models using markov chain monte carlo methods. The book
highlights key issues in modeling polytomous data. Binary data is
presented as a special case of polytomous. There are solid examples
for both.
Latent class models introduces two class models as a
latent trait model and then generalizes to k classes. The book then
discusses maximum likelihood estimation, standard errors and
posterior analysis. Models for manifest variables of mixed type show
how modeling of data sets that contain both continuous and
binary/polytomous data could be done within the generalized
framework.
The final chapters deals with relationships between
latent variables such as correlation between variables. The focus is
on confirmatory factor analysis and linear structural relations
modeling. Related techniques for investigating dependency including
several ways of looking at Principal Component Analysis and
estimation of correlations and regression between latent variables
are covered.
Software appendix provides helpful pointers to
the computer packages that have made many of the methods discussed
practical. The references cover key and up to date academic research
behind the techniques.
Statistical techniques to study the
nature and interpretation of a latent variable should be highly
useful for researchers and practitioners across several fields. The
third edition of this book is comprehensive and provides a solid
foundation for understanding these techniques, and is strongly
recommended.