*2.5. Statistical Analysis*

The data were analyzed using the statistical software IBM SPSS Statistics 21 (IBM Corp, Armonk, NY, USA). Initially, a descriptive analysis was performed with variables at the individual level, which allowed the sample to be characterized for socioeconomic, health, anthropometric, nutritional, and vitamin D statuses. Data on serum levels of 25(OH)D and dietary intake were checked for missing data and outliers that could hinder a multivariate analysis. In addition, variables were excluded from further analysis when they had more than 25% of missing data and outliers, identified under a multivariate perspective through measurement of the Mahalanobis distance (D2)—distance greater than 3.0. There was no record of sample losses and multivariate outliers after adjustment of intrapersonal variability, interpersonal variability, and energy from dietary intake. The normality was checked by using the Kolmogorov–Smirnov test. Subsequently, the variables color, body mass index, cognition, functional status, and seasons of the year were dichotomized according to the predominance of elderly individuals in groups and/or clinical characteristics. Subsequently, to check for statistical associations, Pearson's chi-square test was run, considering the outcome variable (25(OH)D) and other variables relative to the theoretical model. The magnitude of the association was assessed using prevalence ratios (PR) with the corresponding confidence intervals (CIs). The significance level α was set at *p* < 0.05 (two-tailed test).
