NotesFAQContact Us
Search Tips
ERIC Number: ED421523
Record Type: Non-Journal
Publication Date: 1998-Apr-14
Pages: 16
Abstractor: N/A
Reference Count: N/A
Modeling the Observation-to-Indicator Ratio Using Logistic Regression: An Example from Factor Analysis.
Nasser, Fadia; Wisenbaker, Joseph; Benson, Jeri
Logistic regression was used for modeling the observation-to-indicator ratio needed for the standard error scree procedure (SEscree) to correctly identify the number of factors existing in generated sample correlation matrices. The created correlation matrices were manipulated along the number of factors (4,6), sample size (250, 500), magnitude of factor loadings (0.5, 0.8), and degree of interfactor correlations (0, 0.4). Consequently, the observation-to-indicator (n/v) and the indicator-to-factor (v/f) ratios were also changed. The results indicate that the optimal n/v ratio for determining the number of factors by the standard error scree procedure depends on the characteristics of the data. A smaller n/v (7:1) ratio was needed when factor loadings were high and a larger ratio (14-22) was needed with low loading, particularly when factors were correlated. In all conditions, the n/v ratio for the SEscree procedure to correctly identify the true number of factors with high probability exceeded the minimum of 5:1 stated in some of the related literature. Furthermore, the use of logistic regression provided a model for analyzing data from complex simulation studies that makes it very easy to communicate otherwise very complicated relationships. (Contains 4 figures, 3 tables, and 28 references.) (Author/SLD)
Publication Type: Reports - Descriptive; Speeches/Meeting Papers
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A