|
Activity Number:
|
31
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, July 29, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Education
|
| Abstract - #309804 |
|
Title:
|
Measurement Error in Factor Analysis: The Question of Structural Validity
|
|
Author(s):
|
Michael K. Lauritzen*+ and Landon Poppleton and Mitchell Harris and Naomi Hunsaker and Robert Bubb and Bruce Brown
|
|
Companies:
|
Brigham Young University and Brigham Young University and Brigham Young University and Brigham Young University and Brigham Young University and Brigham Young University
|
|
Address:
|
1122 SWKT, Provo, UT, 84602,
|
|
Keywords:
|
Factor analysis ; Structural integrity ; Reliability ; Monte Carlo simulation
|
|
Abstract:
|
The central claim of factor analysis is that it can identify underlying source variables from the correlational pattern in a larger set of surface variables. Monte Carlo simulation is used to create surface variables from a systematic underlying structure, either two-dimensional or three-dimensional, either clustered or Toeplitz, with one of six levels of measurement error added to the data. Quantitative and also graphical methods are used to assess the extent to which structural integrity is maintained for each type of data at each level of measurement error. The watershed value is a reliability of .50. Below that level the structural integrity of the solution deteriorates rapidly. Interestingly, in a survey of papers in two top psychology journals that used factor analysis, not one of the forty-three studies in the sample had adequate reliability levels to provide structural integrity.
|
