Location-scale models are a particularly useful class of models for their ability to simultaneously model the intraindividual variability through the scale sub-model and model the mean outcome through the location sub-model. This thesis extended methods to enhance the interpretability of mixed-effects location scale (MELS) models and address the need for influence analysis and subgrouping techniques within location-scale models. First, due to the relatively complicated structure of MELS models, there are challenges associated with integrating the interpretations of random effects and fixed effects. Researchers often have difficulty comparing the roles that fixed effects and random effects, which represent effects of covariates and between-subject (BS) heterogeneity, respectively, play in the models. Therefore, in the first study, we addressed this problem and enhanced the interpretation of MELS models by introducing R-squared measures as standardized effect size measures for both the location sub-model and the scale sub-model of MELS models. The methodology aligns with the increasing emphasis on reporting actual sizes of effects in addition to or even instead of p-values. Second, as healthcare continues to shift towards personalized approaches, the demand for robust statistical tools to identify individuals with distinctive health behaviors and outcomes is on the rise. In this context, MELS models are valuable instruments for discerning differences among individuals. By developing an influence analysis method framework using MELS models, we aim to enhance our capacity to pinpoint and understand the impact of individual behavior on health conditions and outcomes, contributing to more effective personalized healthcare strategies. In the second study, the proposed approach has successfully demonstrated the use of MELS models in detecting abnormal subjects in intensive longitudinal data that cannot be detected using existing methods based on regular MRMs. We also showcased how this method facilitates a comprehensive understanding of subject heterogeneity in the association between sleep quality and learning goal achievement using data from a published study. Last but not least, the trajectory in intraindividual variability can be an important indicator of treatment success in many research fields, including psychology, behavioral sciences, and clinical developments, to name a few. However, there lacks available methodology to distinguish distinct groups in intraindividual variability trajectory. We proposed a Bayesian location-scale model with latent classes in both mean trajectory and intraindividual variability trajectory to fill this gap. Our developed method, powered by Stan programs, stands as a user-friendly and widely applicable tool. It seamlessly integrates with multiple programming languages, including R, Python, and Stata, making it accessible to researchers, even those without extensive experience in Bayesian modeling and coding. Codes to implement methods described in this thesis are included in the supplementary files.