*2.4. Analyses*

To analyze cross-sectional data, linear regression analyses of the first sound localization test (*n* = 18; two children were not possible to assess) were performed with EI (range = 0.31–1.0) as dependent variable and time since activation of BiCI (i.e., auditory experience, range = 0.03–1.7 years) as the independent variable. To account for betweenand within-subject variability despite missing data points, a linear mixed model was constructed, with the EI as dependent variable and time since activation of BiCI, age at implantation, and the number of obtained responses in a localization test as fixed effects. A random intercept for subjects was included in the model and interaction terms between fixed effects and random intercepts were evaluated. Selection of a final model was guided by minimizing the Aikaike information criterion. The slope of the regression line was statistically compared to slopes obtained in children with normal hearing [28] by an analysis of covariance, and qualitatively to older children with cochlear implants [23]. Statistical analyses were performed using Statistica version 13 (Statsoft, Inc., Tulsa, OK, USA) and R Version 3.4.2 (R Foundation for Statistical Computing, Austria)

Test reliability was computed by dividing each test into two parts and comparing the EI between part 1 (test) and part 2 (retest) [28]. The statistical reliability of the localization test was then quantified by analysis of the variability in test–retest differences and by estimation of the variance in EI for a single SLA measurement (see Equation (10) in [28] for the variance estimate).
