|
Activity Number:
|
149
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2008 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #300749 |
|
Title:
|
Factorial Invariance and Robustness to Low Variability: Maximum Likelihood Factor Analysis vs. Correlation Constraint Analysis
|
|
Author(s):
|
Rochelle E. Tractenberg*+
|
|
Companies:
|
Georgetown University
|
|
Address:
|
Dept of Neurology, Suite 207 Building D, Washington , DC, 20057,
|
|
Keywords:
|
factor analysis ; tetrad ; causal modeling ; factorial invariance ; latent structure
|
|
Abstract:
|
This poster describes a planned simulation study comparing methods to uncover latent structure. Exploratory factor analysis (EFA) will be contrasted with the more explicitly causal correlation constraint analysis (CCA). The project will compare CCA and EFA in terms of accuracy, invariance, and sensitivity to variability within 100 simulated samples built to specifications and repeated with high and low levels of variability. The study will test whether: 1. CCA is more sensitive to lower levels of variability than EFA. 2. CCA results are more factorially-invariant than EFA results. The proposal will determine if CCA correctly recovers the 'true' latent variables and structure from simulated observed data more consistently than EFA ("accuracy"), recover the same 'true' model 95% of the time ("invariance"). Replication of results in samples with high/low variance will support sensitivity.
|
