Structural equation modeling compared with ordinary least squares in simulations and life insurers’ data

Structural equation model (SEM) is a general approach to analyze multivariate data. It is a relatively comprehensive model and combines useful characteristics from many statistical approaches, thus enjoys a variety of advantages when dealing complex relationships. This report gives a brief introduction to SEM, focusing especially the comparison of SEM and OLS regression. A simple tutorial of how to apply SEM is also included with the introduction and comparison. SEM can be roughly seen as OLS regression added with features such as simultaneous estimation, latent factors and autocorrelation. Therefore, SEM enjoys a variety of advantages over OLS regression. However, it is not always the case that SEM will be the optimal choice. The biggest concern is the complexity of SEM, for simpler model will be preferable for researchers when the fitness is similar. Two simulation cases, one requires special features of SEM and one satisfies assumptions of OLS regression, are applied to illustrate the choice between SEM and OLS regression. A study using data from US life insurers in the year 1994 serves as a further illustration. The conclusion is when special features of SEM is required, SEM fits better and will be the better choice, while when OLS regression assumptions are satisfied, SEM and OLS regression will fit equally well, considering the complexity of SEM, OLS regression will be the better choice.

Name thesis:degree_name

Master of Science in Statistics

Level thesis:degree_level

Masters

Discipline thesis:degree_discipline

Statistics

Grantor

The University of Texas at Austin

Year dc:date.issued

2013

Author dc:creator

• Xiao, Xuan, active 2013

Advisor dc:contributor.advisor

• Sager, Thomas W.

• Structural equation modeling
• Ordinary least squares
• Comparison

Language dc:language.iso

en_US

Handle dc:identifier.uri

http://hdl.handle.net/2152/22516

OAI identifier oai:identifier

oai:repositories.lib.utexas.edu:2152/22516