Categorization and Measurement Quality. The Choice Between Pearson and Polychoric Correlations
Coenders, G. & Saris, W. E.
Saris, W. E. & Münnich, Á. (Eds.) The Multitrait-Multimethod Approach to Evaluate Measurement Instruments. Eötvös University Press, Budapest (1995): 125-144.
This chapter first reviews the major consequences of ordinal measurement in Confirmator y Factor Analysis (CFA) models and introduces the reader to the Polychoric correlation coefficient, a measure of association designed to avoid these consequences. Next, it presents the results of a Monte Carlo experiment in which the performance of the es timates of a CFA model based on Pearson and Polychoric correlations is compared under different distributional settings. The chapter concludes that, in general, it does not matter too much which measure of association one uses, as long as one is aware that factor loadings should be interpreted differently, depending on whether Pearson or Polychoric correlations are analysed.
Full text print proofs (*.pdf file)
Back to main page