The Use of Exploratory and Confirmatory Factor Analyses: A Document Analysis

Creative Commons License

Koyuncu I., KILIÇ A. F.

EGITIM VE BILIM-EDUCATION AND SCIENCE, vol.44, no.198, pp.361-388, 2019 (SSCI) identifier identifier

  • Publication Type: Article / Article
  • Volume: 44 Issue: 198
  • Publication Date: 2019
  • Doi Number: 10.15390/eb.2019.7665
  • Journal Indexes: Social Sciences Citation Index (SSCI), Scopus, TR DİZİN (ULAKBİM)
  • Page Numbers: pp.361-388
  • Hacettepe University Affiliated: No


This paper aims to review the scale development research published in Turkey between 2006 and 2016 with regard to their processes of exploratory (EFA) and confirmatory factor analysis (CFA). Within this scope, the distribution of the studies according to years and factor analyses, the extent to which their hypotheses came through for both analysis methods, and the distribution of EFA and CFA results according to the published papers were investigated. In this way, we aimed to infer significant results on the practical application of EFA and CFA which are frequently referred in theory. Hence, the present research incorporates a descriptive document analysis on 131 scale development studies published in the fields of education and social sciences in the journals indexed in The Scientific and Technological Research Council of Turkey (TUBITAK) TR Dizin data base between 2006 and 2016. Frequency tables, column charts, histogram, line charts, and measures of central tendency were used to analyse the data. More than 70% of 131 scale development studies were carried out after 2013. The maximum number of studies was published in 2015, and the minimum number in 2008. Only EFA was used in 59 articles, only CFA was used in 4 articles, and both analysis methods were employed in 58 articles. The remaining 10 papers used none of the analyses and included only item analyses. The average sample sizes were 395 participants for EFA and 529 participants for CFA. Sample sizes were accordingly adequate in terms of the factor analyses. EFA and CFA were conducted over the same sample in 36 papers, whereas the analyses were based on different samples in 22 ones. In terms of estimation, 65% of 117 EFA using papers employed Principle Component Analysis (PCA), 2% used Maximum Likelihood Estimation (MLE), 1% used Parallel Analysis (PA), and 3% used Principle Axis Factoring (PAF), while 29% did not specify their method of estimation. The hypotheses of EFA analyses were examined in 93% of the papers. Scree plots were taken into consideration when determing the number of factors in 50% of the papers. Varimax rotation method was used in 67% of the EFA using papers. The total variance explained was 53.05%. The analytical hypotheses were not examined in 87% of the 62 studies conducted with CFA; 51 of them included track diagrams and 52 included factor loadings. 90% of the papers did not give any information on parameter estimation method. All indices, except the Parsimony Goodness of Fit Index (PGFI), were acceptable as model fit indices in the papers. The results from the papers examined were evaluated in the light of the relevant literature in the last section of the present paper, and some suggestions were made in reference to both theory and practice.