All statistical analyses will be performed using Statistical Package for the Social Sciences (SPSS, Version 25.0) software. Time will be coded as 0, 1, 2, and 4 to reflect surveys given at baseline, three months post-transplant, six months post-transplant, and twelve months post-transplant to account for the unequal amount of time between timepoints. Preliminary analyses will be completed to identify any missing data or outliers.
The data will be evaluated for all assumptions related to the planned analyses including linear mixed models, Pearson r correlations, and Cox proportional hazard models. The following assumptions will be tested for linear mixed models. First, linearity will be tested by plotting the model residuals by the predictors (social support and treatment adherence). If the pattern in the plot is random, the assumption is met. However, if the pattern is non-random and follows a trend, predictors will be transformed. Next, homoscedasticity will be examined using scatter plots of the residual values. The assumption is met if the points are equally distributed linearly. QQ plots will test the assumption of normally distributed residuals in the model. The assumption is met if the residuals follow a linear pattern and do not deviate from the expected normal line. If the assumption is not met, log/ln transformations will be used to improve the normality of the data.
The data will be evaluated for a linear relationship between the social support and treatment adherence trajectories by plotting a scatterplot for the slopes of both trajectories. The assumption is met if there is a straight-line relationship between the two variables.
The proportional hazards assumption will be tested for Cox proportional hazard models. The proportional hazards assumption posits that the hazard ratio between two groups must remain constant over time and will be examined using Kaplan-Meier curves. If the curves cross, or the overall test of proportional hazards is significant, the proportional hazards assumption has been violated.
Gender and age may be included as possible covariates in all analyses, given that they may explain variations in social support, treatment adherence, and survival. To determine if gender or age should be included as a covariate, bivariate correlation analyses will be performed between both potential covariates, gender and age, and the outcome variables, social support, treatment adherence, and longevity. If the possible covariates are correlated with the dependent variables at a large Pearson correlation greater than .5, they will be included in all analyses as covariates. However, if the Pearson correlation is not large, covariates will not be included in analyses.