Crystal Smit

Chapter 4 78 RESULTS Preliminary Analyses Descriptive Statistics and Randomization Check Descriptive statistics showed that on average children consumed 2.99 ( SD = 1.70) glasses of water and .57 ( SD = .58) glasses of SSB a day at baseline (T1). The means and standard deviations for all study variables across the conditions are summarized in Table 4.1. Table 4.1 Means and standard deviations for all study variables across the conditions per assessment SNI ( n = 149) Active Control ( n = 164) Control ( n = 138) M SD M SD M SD Water consumption T1 3.03 1.89 3.12 1.58 2.79 1.62 Water consumption T2 3.12 1.99 3.12 1.66 2.76 1.97 Water consumption T3 3.07 2.07 2.29 1.87 2.62 2.00 SSB consumption T1 a .67 .72 .49 .47 .55 .58 SSB consumption T2 .64 .79 .45 .42 .73 1.06 SSB consumption T3 .61 .69 .57 .59 .81 1.21 Descriptive norms T1 3.62 1.03 3.55 1.04 3.74 1.11 Injunctive norms T1 3.69 1.65 3.73 1.65 2.97 1.77 Note. SNI, social network intervention; M, mean; SD , standard deviation; T1, Time 1; T2, Time 2; T3, Time 3; a the three conditions differed significantly on this variable ( p < .05). To check whether there were initial mean-level differences between the three conditions on the outcome variables (i.e., water and SSB consumption), one-way ANOVA were conducted. The analyses yielded statistically significant differences at baseline (T1) between conditions for SSB, F (2,435) = 3.57, p = .029, but not for water consumption, F (2,435) = 1.38, p = .252 (see Table 4.1). This indicated that the randomization was not successful for SSB consumption; it is therefore essential to account for these initial differences between conditions to avoid interpreting regression to the mean effects (i.e., groups that have low mean scores are more likely to increase; Barnett, Van Der Pols, & Dobson, 2005). To account for these initial differences, we included baseline consumption behavior (T1) as a predictor of consumption behavior at T2 and T3 in the structural path models.