Similar commonly used methods of instrument structure examination include
principal component analysis (PCA) and confirmatory factor analysis (CFA). The
main difference between CFA and Rasch analysis on the one hand, and PCA on the
other hand, is that the first two investigate how well the data fit a hypothesized
model, while the latter is a method of data reduction. That is, whereas Rasch analysis
provides fit statistics with respect to how well an item fits into a dimension of a
hypothesized trait structure and CFA determines the factor loadings of an item with
respect to different dimensions, with PCA high-correlating items can be identified
with the purpose of removing unnecessary items. When developing a questionnaire
to survey a theoretically well-defined trait such as views of NOS, Rasch analysis and
CFA would be the more appropriate methods to investigate whether the questionnaire
actually represents the theoretically hypothesized structure. Although the
results obtained from Rasch analysis and CFA can principally be converted into
each other (Edwards & Wirth, 2009), the procedures are based on different assumptions
regarding the original data: CFA requires the original data to be normally
distributed on an interval scale, whereas polytomous Rasch analysis only requires the
data to be ordinal. Thus, although CFA is commonly used to analyze Likert-based
questionnaires (assuming Likert-type questions are discrete representations of a
continuous interval scale), Rasch analysis would be the more appropriate method in
this respect, which is why we chose Rasch analysis for the investigation of the instrument
developed by Lombrozo et al. (2008).