The statistical analysis was performed separately for each laboratory (A, B) and the two methods (HI, SNT) for the logarithmic titres in a mixed-linear model for repeated measurements.
The primary outcome was compared using a mixed linear model, adjusting for the period effect.
The primary study end point was analyzed via a mixed linear model, adjusting for period, treatment sequence (carryover), and treatment effect (fixed factors), with patient indicator as a random intercept and PR at baseline as a covariate.
Changes from EV1 to EV2 and EV3 were tested within groups and compared between groups using a mixed linear model adjusting for season of the year and type of surface (carpet vs. hard).
EMG analysis was computed with Box-Cox transformation to normalize data and increase applicability and usefulness of the estimation procedure on structured data throughout a mixed linear model adjusted for each muscle.
For each trait, a mixed linear model (MLM) accounting for kinship (K matrix) and population structure (Q matrix) was used.
The QK model, a mixed linear model (MLM) adjusting for both population structure and genetic relatedness between genotypes, implemented by TASSEL was used for association analysis.
We used a mixed linear model to test for association between SNP variation in 35 selection-candidate genes and trait variation in teosinte [11], [12].
In addition, a mixed linear model was fitted for each gene to estimate the within-subject correlation.
