The GBR-wide trend in coral cover was estimated using a repeated-measures design with a fixed linear time parameter and random time and intercept effects for reefs.
Next, differences in sniff response between pleasant and unpleasant odors and specific sniff parameters as a function of group (ASD/TD) were estimated using a multivariate repeated-measures ANOVA with conditions of sniff parameter (mean airflow, airflow peak, sniff duration, and sniff volume), odorant valence (pleasant or unpleasant), and group (ASD or TD).
The proportion of reversals and their 95% confidence intervals were estimated using a GEE model for repeated binary data, assuming a binomial distribution and log link, and accounting for subject as a repetition factor.
For Z scores, change in body temperature, and serum concentration of CRP, the adjusted difference of treatment mean and its 95% CIs, will be estimated using a mixed effects model, for repeated measures.
To assess non-inferiority for the primary endpoint, the upper bound of the two-sided 95% CI of the least squares (LS) mean difference, estimated using a mixed-effect model with repeated measures approach, was compared with a pre-defined non-inferiority margin (<0.40 % HbA1c; <4.4 mmol/mol).
Habitual food intake was estimated using a Food Questionnaire (FQ) which was repeated within two weeks.
The impact of skin toxicity grade (0/1 or 3/4) on changes from baseline in health state index, overall health rating and global health status scores was estimated using a linear mixed effects model for repeated measures.
The treatment difference was estimated using a likelihood-based mixed model for repeated measures (MMRM) analysis with baseline HbA1c as a covariate; treatment, country, week of visit and the treatment-by-week interaction as fixed effects; and patient and error as random effects.
