*2.9. Statistical Analysis*

Information as to whether recruits completed academy and graduated, or did not and were separated, were provided by training staff from the LEA. Recruits were then split into four groups based on the information provided by LEA staff: Those that graduated (GRAD), and those that separated for personal reasons (SEPPR), PT failures or injury (SEPFI), or due to academic or scenario failures (SEPAS). Separation due to PT failures and injury were initially two separate groups. However, they were combined into one group, due to less fit recruits being more likely to ge<sup>t</sup> injured during academy [47], and because injuries often led to a recruit not completing the required number of PT sessions (which then resulted in academy separation). Sexes were combined within these groups, as all recruits need to attain the same standards to graduate academy, regardless of sex. This approach has been used in previous research [20,22,23,27,28].

Statistical analyses were computed using the Statistics Package for Social Sciences (Version 25.0; IBM Corporation, New York, USA) and Microsoft Excel (Microsoft Office Professional Plus 2016, Microsoft Corporation, Washington, WA, USA). Descriptive statistics (mean ± standard deviation [SD]) were calculated for each test parameter. A one-way analysis of variance (ANOVA), with Bonferroni post hoc for multiple comparisons, was used to calculate any performance differences in the fitness

tests between the four groups. Significance was set at *p* < 0.05 a priori. Similar to previous research [23], effect sizes (*d*) were also calculated for the between-group comparisons, where the difference between the means was divided by the pooled SD [48]. In accordance with Hopkins [49], a *d* less than 0.2 was considered a trivial effect; 0.2 to 0.6 a small effect; 0.6 to 1.2 a moderate effect; 1.2 to 2.0 a large effect; 2.0 to 4.0 a very large effect; and 4.0 and above an extremely large effect. Multiple stepwise linear regression was used to determine whether age, height, body mass, or the physical fitness tests predicted graduation or reasons for separation in the recruits. As group inclusion was a categorical variable within SPSS, the data were recoded into dummy variables to provide dichotomous values (1 = group inclusion for either GRAD, SEPPR, SEPFI, or SEPAS; 0 = all other groups). Thus, GRAD, SEPPR, SEPFI, or SEPAS each acted as a dependent variable [6].
