Statistical analyses

Different statistical approaches were used to address the research questions. For research question 1, a latent variable approach was chosen, for research question 2 a repeated measures analysis of variance (ANOVA) was conducted, and for research question 3 both a repeated measures ANOVA and a latent variable approach were implemented. The analyses were conducted using IBM SPSS (Version 26) and MPlus 8 (Muthén & Muthén, 1998–2017). The latent variables were used to minimize measurement error. To estimate the structural equation model (SEM), the full information maximum likelihood method (ML) was used. To evaluate the goodness of fit for the model, we used the chi-square test statistic and several commonly recommended descriptive measures of model fit (Hu & Bentler, 1998): the standardized root mean squared residual (SRMR), the root mean square error of approximation (RMSEA), the comparative fit index (CFI), and the Tucker-Lewis Index (TLI). Models were evaluated as fitting well when the baseline fit indices (SRMR and RMSEA) were < 0.08 and the incremental fit index (CFI) was ≥ 0.95.

Analyses of missing data lost due to attrition showed that the overall percentage of missing values for the sample was 2.31% with 25 total cases containing missing data (in one to four cells). Little’s missing completely at random (MCAR) test was not significant, χ²= 161.92, df = 177, p = .785, suggesting no definitive pattern to the missing data. In order to prevent a loss of power (Graham, Olchowski, & Gilreath, 2007), missing data was addressed via multiple imputation (MI). With this approach, several new data sets were created from original data, in which missing values were replaced by probable values. Overall, 10 imputations were conducted in Mplus. The SEMs were conducted on the imputed and non-imputed data sets. Comparisons yielded no significant differences in the interpretation of results, indicating that magnitude, directions, and statistical significance of effects were consistent. Therefore, all analyses reported are conducted using the imputed data sets to maximize the sample size available.