Mediation Analyses for Carbohydrate Counting
As participants in the intervention condition showed significant
improvement over time for carbohydrate counting, a mediation analysis
(Hayes, 2013) was conducted to assess
whether the integrated model variables mediated the impact of the
intervention on carbohydrate counting. In addition to pre-intervention
carbohydrate counting behavior, all potential mediators, along with
Condition, were entered into the model. The direct effect of Condition
on behavior was significant, B=0.54, SE=0.14, p<0.01 .
Bootstrapping analyses resulted in the total significant mediated
(indirect) effect as B=0.26, SE=0.12, CI=0.01 to 0.51 . The
indirect effect was significant only via planning, B=0.10,
SE=0.06, CI=0.01 to 0.26 . Further, inspection revealed that, after the
inclusion of potential mediators (integrated model variables), the
direct effect of the intervention on carbohydrate counting behaviour
remained significant, B=0.37, SE=0.16, CI=0.03 to 0.71,indicating a partial mediation effect.