Statistical software version 12

We used descriptive statistics to determine means and frequencies of demographics and food purchasing patterns and dietary intake. To model the direct association between the food purchasing patterns and dietary intake, multiple linear regression models were used, controlling for race, age, and sex. To model the mediation effect of the home, school, and consumer food environments on the relationship between food purchasing patterns and dietary intake, a seemingly unrelated regression model (sureg) was used to quantify the indirect effects of all three potentially mediating variables while controlling for race, age, and sex. The seemingly unrelated regression model utilizes the product of the coefficients method, finding the indirect effect of the mediating variables by multiplying the regression coefficients from the independent variable and the mediating variable, with the mediating variable on the dependent variable. This model was used due to the categorical nature of the independent variable, food purchasing patterns. All analyses were coded with Stata data analysis and statistical software version 12.0 (StataCorp LLC., College Station, TX, USA).

Chi-squared, logistic regression and correlation (Pearson and Spearman) analyses were completed using Stata data analysis and statistical software (Version 12). Prior to statistical analysis an initial model was created with ‘sometimes’ response options excluded. This allowed direct comparison between those answering definitively ‘yes’ or ‘no’ (Model 1). In the interest of completeness, survey response options were also dichotomised yielding two further datasets for analysis i.e. Model 2 = Y + S versus N and Model 3 = Y versus S + N. A total of five herd classification independent variables (i.e. region, calving-season, enterprise type, herd size, bioexclusion status) were used to examine key influences on JD risk variables. As a first step, a univariable (Pearson’s Chi-squared) analysis was completed. Independent variables recording P ≤ 0.15 were included in logistic regression models (1, 2 and 3). A manual backwards elimination with a forward step was applied to each model with significant variables (P ≤ 0.05 chosen as accepted significance level) retained in the final model. Pearson’s correlation was used to check for co-linearity across independent variables. Spearman’s rank correlation (r_s) was performed to examine relationships between dependent variables (JD survey questions) with r_s values of >0.3 reported. Biosecurity variables were not statistically analysed.