Our cohorts consist of STEM students (N = 2374) from different engineering courses at Ruhr-Universität Bochum in Germany, typically predominantly males, some with insufficient background in mathematics or non-native speakers of German. The data was analysed using various statistic methods, e.g. reliability measurement and confirmatory factor analysis. Our findings show that about half of the original items (36 out of 69) are sufficient, reliability holds (Cronbach’s α > 0.7) and more variance is explained (56.17 % as compared to 45.65 %). Most content-related changes occurring when eliminating so many items survive critical scrutiny.