Plan of analyses
1) Descriptive statistics: scale means and standard deviations were calculated within the total sample and within subgroups (gender, prior study track). Means and correlations were calculated for those students for which variables were available. Significance of differences in mean scores by gender and prior study track was investigated by means of independent t-tests, post-hoc tests and calculation of effect sizes. Cut-off criteria of Cohen's d [44] were used, in which d = .2 is indicative of a small effect while d = .5 and d = .8 represent a medium and large effect respectively. 2) Pearson correlations were calculated to explore the relationship within and between the motivational variables and variables related to prior and early academic achievement. Criteria of Cohen [44] were applied to interpret the strength of the correlation patterns, in which r>.10 and . 30 and .50 for a strong correlation.3) Cluster analyses on autonomous motivation and academic self-concept were carried out to distinguish motivational profiles among student groups. A two-step procedure was used, with a hierarchical clustering procedure (Ward's method) to determine the optimal cluster number and initial seeds for a second non-hierarchical k-means clustering procedure, as described in [13]. A cluster solution will be retained in the analyses if the variance in the constituting dimensions is above the 50% threshold [43]. To examine the stability of a cluster solution, the double-split cross-validation procedure was used [13]. 4) Variance analyses and calculation of effect sizes was carried out to explore the relationship between the obtained cluster solution, early academic achievement and gender.