The use of polychoric and Pearson correlation matrices in the determination of construct validity of Likert type scales


Ozdemir H. F., TORAMAN Ç., KUTLU Ö.

TURKISH JOURNAL OF EDUCATION, vol.8, no.3, pp.180-195, 2019 (ESCI) identifier

  • Publication Type: Article / Article
  • Volume: 8 Issue: 3
  • Publication Date: 2019
  • Doi Number: 10.19128/turje.519235
  • Journal Name: TURKISH JOURNAL OF EDUCATION
  • Journal Indexes: Emerging Sources Citation Index (ESCI), TR DİZİN (ULAKBİM)
  • Page Numbers: pp.180-195
  • Keywords: Factor analysis, Polychoric correlation matrix, Pearson correlation matrix, Likert type scales, Construct validity
  • Çanakkale Onsekiz Mart University Affiliated: Yes

Abstract

No matter how strong the theoretical infrastructure of a study is, if the measurement instruments do not have the necessary psychometric qualities, there will be a question of trust in interpreting the findings, and it will be inevitable to make wrong decisions with the results. One of the important steps in scale development/adaptation studies is to provide evidence of the experimental validity. In order to reveal evidence of construct validity of Likert type scales, to identify factor structures, to confirm previously predicted structures, factor analysis is used. The primary issue to be examined is the level of measurement of the variable and one of the leading decisions that must be taken is which relation matrix will be used. This descriptive research is based on the effects of using Pearson or polychoric correlation matrix in the factor analysis. It is determined that items showed different "item-total correlations", "loading values" and "correlation coefficients", different factor numbers emerged, different items were removed out of the scale, confirmation status of the structure has changed.