Multiple regression analyses were planned using up to three predictor variables and one dependent variable per analysis for each self-care component. Each self-care component was treated as a separate dependent variable per the recommendations of Toobert et al. (2000). The sample size was estimated a priori using GPOWER, a program that calculates the estimated power of a statistical test given various sample sizes (Faul & Erdfelder,1992). With GPOWER , we determined that a sample size of 85 would provide 80% power to detect a medium effect of .30, for two-tailed tests for calculating Pearson’s correlations and a medium effect size of .15 of 3 predictors on the value of F to detect changes in a specific self-care behavior (diet, physical activity, monitoring,
pill, or insulin adherence) at an alpha of .05.