*2.7. Statistical Analysis*

Statistical analyses were performed in Matlab. First, for subjects with normal scapulae (i.e., trauma patients), we performed simple and stepwise multiple linear regressions to examine the correlation among all six acromion landmarks (AA and AC coordinates) and the GRA. We also evaluated the correlation among each of the four acromion angles (APA, ATA, ALA, and AXA) and the GRA. The quality of the regression was quantified by the root mean square error (RMSE), the coefficient of determination (R2), and its *p*-value. We further performed a receiver operating characteristic (ROC) curve analysis to determine which critical GRA and associated morphological acromion parameter better identified the two groups (i.e., low vs. high GRA), using the area under the curve (AUC) with the Youden index. The normality of the measurement data was verified by a Shapiro–Wilk test. As an additional analysis, differences between the normal and pathological patient groups were tested by an unpaired two-tailed Student's *t*-test, and the effect size evaluated with Cohen's d. We also assessed the dependence on patient demographics such as gender and age, and *p* < 0.05 was considered statistically significant.
