Lakhasly

Weighted descriptive statistics were calculated for the dependent variable (eating concerns), two independent variables (social media volume and frequency), and each of the seven covariates.

Exploratory Factor Analysis using Principal Components Analysis (PCA) with varimax rotation was performed to assess the underlying structure of the eating concerns items, and Cronbach’s α was used to examine the internal consistency reliability of the eating concerns items.

Chi-square tests were used to determine bivariable associations between each of the independent variables and covariates and the dependent variable. Additionally, bivariable associations between each of the covariates and independent variables were assessed using chi-square tests.

After confirming that the proportional odds assumption was met for each analysis, ordered logistic regression was used to assess bivariable and multivariable associations between each independent variable and the dependent variable, which was an ordered categorical variable. It was decided a priori to include all covariates in multivariable models. Additionally, tests for interaction effects between each independent variable and all covariates were performed. The presence of an overall linear trend between each ordered categorical independent variable and the dependent variable was tested using an established method.32

Two sets of sensitivity analyses were conducted to confirm the robustness of the results. First, auxiliary analyses were conducted that modeled the dependent variable (eating concerns) as dichotomous and in quartiles. Second, analyses were conducted that modeled the independent variables (social media volume and frequency) as continuous.

Study-specific post-stratification weights provided by GfK were used to perform all descriptive statistics and analyses. These weights were computed to adjust for non-response as well as non-coverage, under-, or over-sampling resulting from the sample design. Statistical analyses were performed with Stata 12.1,33 and two-tailed P-values < 0.05 were considered to be significant.

Weighted descriptive statistics were calculated for the dependent variable (eating concerns), two independent variables (social media volume and frequency), and each of the seven covariates.

Exploratory Factor Analysis using Principal Components Analysis (PCA) with varimax rotation was performed to assess the underlying structure of the eating concerns items, and Cronbach’s α was used to examine the internal consistency reliability of the eating concerns items.

Chi-square tests were used to determine bivariable associations between each of the independent variables and covariates and the dependent variable. Additionally, bivariable associations between each of the covariates and independent variables were assessed using chi-square tests.

After confirming that the proportional odds assumption was met for each analysis, ordered logistic regression was used to assess bivariable and multivariable associations between each independent variable and the dependent variable, which was an ordered categorical variable. It was decided a priori to include all covariates in multivariable models. Additionally, tests for interaction effects between each independent variable and all covariates were performed. The presence of an overall linear trend between each ordered categorical independent variable and the dependent variable was tested using an established method.32

Two sets of sensitivity analyses were conducted to confirm the robustness of the results. First, auxiliary analyses were conducted that modeled the dependent variable (eating concerns) as dichotomous and in quartiles. Second, analyses were conducted that modeled the independent variables (social media volume and frequency) as continuous.

Study-specific post-stratification weights provided by GfK were used to perform all descriptive statistics and analyses. These weights were computed to adjust for non-response as well as non-coverage, under-, or over-sampling resulting from the sample design. Statistical analyses were performed with Stata 12.1,33 and two-tailed P-values < 0.05 were considered to be significant.