平行分析於單因素二分變項之應用性
Date Issued
2003
Date
2003
Author(s)
DOI
912413H002015
Abstract
Determining the number of factors is a
crucial step in factor analysis. The present
Monte Carlo research investigated the
performance of parallel analysis as applied to
unidimensional binary data. The simulation
study employed an eight-variable
single-factor model and manipulated sample
size (50, 100, 200, 500, 1000), size of factor
loadings (.45, .70, .90), response ratio on two
categories (50/50, 60/40, 70/30, 80/20,
90/10), and type of correlation used (phi
correlation and tetrachoric correlation) for
analysis. The results indicated that parallel
analysis suggested very accurate number of
factors. The proportion of correctly
identifying the number of factor increased as
sample size and factor loading increased, and
as the percentages of responses on two
categories became close. Parallel analysis
that used 95th and 99th percentile of the
random data eigenvalues as the basis for
comparison yielded higher correct rate then
using mean eigenvalues as the criterion.
The performance of phi correlations and
tetrachoric correlations had only minor
differences. Overall, the size of the factor
loading had the most significant effect on the
performance of parallel analysis with binary
3 data.
Subjects
factor analysis
parallel analysis
dichotomous variables
unidimensionality
simulations
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
912413H002015.pdf
Size
154.33 KB
Format
Adobe PDF
Checksum
(MD5):8f7e6c5411491942d75ffc342cf6da98