平行分析於李克式量表之評估(1/2)
Date Issued
2005
Date
2005
Author(s)
DOI
932413H002022
Abstract
Parallel analysis has been shown to be superior to other methods for
determining the number of factors in exploratory factor analysis. The present study
was designed to examine the performance of parallel analysis on orthogonal-factor
models with indicators measured on Likert-type rating scales. The simulation study
manipulated model complexity as indicated by the size of trivial loadings (0 or 0.1),
number of factors (3 or 6), number of indicators per factor (4 or 8), size of major
factor loadings (0.4, 0.6, 0.8), sample size (100, 200, 500), number of response
categories on the Likert scale (3, 5, 7), and distribution of the Likert-type variables
(skewness of 0, 1, or 2). Parallel analysis was performed using the mean, the 95th
percentile, and the 99th percentile of the eigenvalues from 1000 random data
correlation matrices. The size of major factor loading played the most significant
role in the behavior of parallel analysis. In general, parallel analysis tended to
suggest the correct number of factors on Likert-type data, especially with high major
factor loadings. However, when the complexity of the factor model increased and
the size of major factor loading decreased, parallel analysis might have difficulty
identifying the correct number of factors.
Subjects
Number of factors
parallel analysis
Likert-type scales
Publisher
臺北市：國立臺灣大學心理學系暨研究所
Type
report
File(s)![Thumbnail Image]()
Loading...
Name
932413H002022.pdf
Size
56.25 KB
Format
Adobe PDF
Checksum
(MD5):554b4982fe43a7c5c142791df8e363e1