Only participants with complete data for all sections were included in the analysis. Mean (standard deviation, SD), median (interquartile range, IQR), and proportions were used to describe participants on key characteristics and outcomes based on data distribution. Comparisons between sociodemographic characteristics were performed using parametric and non-parametric tests to compare behaviors before and during COVID-19. Multicollinearity was assessed with variable inflation factors (VIF). We used multiple linear regressions with a residualized change score approach [13,14] to investigate the sociodemographic predictors (independent variables) of changes in physical activity, recreational screen time, sleep duration, and sleep quality (dependent variables) during the early stages of the pandemic in Chile. This approach provides robust estimates by eliminating auto-correlated errors and regression towards the mean, which often makes it preferable to the simple change score approach [13]. First, we regressed the standardized score during the COVID-19 pandemic on the standardized scores before the COVID-19 pandemic for each of the behaviors assessed in this study. The residualized change score (i.e., trend) for each behavior was then estimated as the average of each participant’s residual score (i.e., the difference between the estimated value and the observed value). A positive residualized change score indicates an increase in the specific behavior from the time before COVID-19 and a negative score indicates a decrease. For each residualized change score, we explored the predictive role of a series of sociodemographic factors a priori thought to have influenced changes in the behaviors herein assessed after ECEC and school closures due to COVID-19. We adjusted all models for age and sex of the child, family income, presence of lockdown, and region. Data preparation and validation was conducted with Stata 15.0 (College Station, TX, USA: StataCorp LLC), while analyses were conducted in R (version 3.5.2) (R Foundation for Statistical Computing, Vienna, Austria). The level of significance was set at usual p < 0.05, two-tailed.