32 Chapter 2 Figure 2.2. Grid showing the prearranged frequency distribution for the Q-sorting process Step 4: Data analysis We used PQMethod version 2.35 to perform factor analysis on the Q-sorts (33). PQMethod is a software program specifically developed for performing by-person (instead of byitem) factor analyses in Q-methodological studies. In line with Watts and Stenner’s recommendations (24), we employed the centroid method of factor analyses, with varimax rotation, complemented with manual rotations. The centroid method leaves researchers “… free to consider any data set from a variety of perspectives, before selecting the rotated solution which they consider to be the most appropriate and theoretically informative” (34). This method suited our aim to include as many students in the factors as possible. Other methods, like principal component analysis, do not offer this freedom as they prescribe on statistical criteria alone which one solution to accept (35). This is also the reason for complementing the varimax rotation with manual rotations, to evaluate if we could add extra students to a factor. Three researchers (JWG, AdlC, RK) evaluated the outcomes of the factor analyses (i.e., factor solutions) and decided on the accepted solution through consensus. Our criteria for accepting a solution were statistical (eigenvalues of >1.00, minimal total explained variance of 35%, and at least 2 Q-sorts per factor), qualitative (corroboration of the factor solution by the post-sorting interview data), and methodological (are the factors coherent, differentiated and recognizable) (24). As a final step, we used the study’s conceptual framework to characterize the profiles. Step 5: Factor interpretation We followed the structured method for factor interpretation provided by Watts and Stenner (24). We started with the calculation of factor arrays (weighted averages of the Q-sorts in a factor, see Table 2.1). Factor arrays show how a prototypical student in a factor would
RkJQdWJsaXNoZXIy MTk4NDMw