| Appendix 2 Details of factor analysesThe Which University? questionnaire comprised three
major sections designed to probe the various influences we expected would affect
applicants’ field of study, course and university preferences. Responses were on a
5-point Likert-type scale. Statistical analyses were used to confirm the conceptual
structure of the instrument and to aid in reporting of the data.
For both the school-leaver and mature-age subgroups, factor analytic techniques were used in turn on each of the three sections of the questionnaire. Principal components analysis (PCA) was initially employed to determine the number of factors present. The factor solution suggested by the PCA was then rotated to increase the interpretability of the extracted factors. Following rotation, the item-composition of each of the factors was determined. The factor structure suggested by the PCA was used as the basis for further analyses in which the more confirmatory one factor congeneric modelling technique was employed. This technique is a specific case of confirmatory factor analysis, in which each of the factors identified in the previous exploratory analyses is tested for construct validity. To illustrate the overall approach, the following section provides details on the analysis of the field of study preferences section for school-leaver respondents. Further details of the analyses are available to interested researchers. Analysis for school-leavers field of study preferences Prior to analysis, the item distributions were examined for normality. Normality checks indicated that the distributions conformed to the assumption of normality. The scree test method and Kaiser’s criterion (i.e. eigenvalues greater than one) were used to determine the number of factors for extraction. The outcome of the principal components extraction above indicated that items could be grouped according to four main constructs. In order to determine the item-composition of each of the factors the initial solution was rotated. Loadings of variables on factors and communalities are shown in Table A1. Variables are presented within each factor and ordered by size of loading. Individual loadings were generally high, as were the associated communality values, and each of the factors was well defined. Further analyses of the 4-factor structure suggested by the PCA above were conducted using the more confirmatory one factor congeneric modelling technique. This technique tests the construct validity of each individual factor. Table A1 Factor loadings and communality values (h2) for exploratory factor analysis using a varimax rotation and principal components extraction
Note: * Loadings > .39 reported Estimated factor scores Estimated factor scores were used to provide a measure of each of the constructs generated. Table A2 details the calculations of each of the factor scores and their corresponding means and standard deviations. The mean scores indicate that the constructs identified in the analyses differed in the degree to which they influenced students’ decisions to study in particular areas. Factor score intercorrelations were also calculated. Generally, these were all significant, suggesting that the factors represent related, yet distinct constructs. Table A2 Estimated factor score means and standard deviations
|