In many psychological and educational applications, it is imperative to obtain valid and reliable score estimates of multilevel processes. For example, in order to assess the quality and characteristics of high impact learning processes, one must compute accurate scores representative of student- and classroom-level constructs. Currently, there are two general approaches to analyzing multilevel latent variable item response: (1) the aggregated approach or single-level factor analysis with cluster robust standard errors and (2) multilevel factor models (e.g., Stapleton, McNeish, et al., 2016; Stapleton, Yang, et al., 2016); however, there is a gap in methodological literature regarding which method produces optimal level-1 and level-2 factor score estimates. An additional complication arises when sources of measurement invariance exist at level-2. Past research has suggested that incorporating background characteristics into scoring models improves factor score estimates (Curran et al., 2016), but this has not been expanded to a multilevel context. Thus, the present study extensively evaluates factor scores extracted from a variety of measurement models, in the presence of measurement invariance at level-2. Both simulation methodology evaluating correlations between factor scores and true scores and an empirical demonstration with educational data were utilized. No single scoring procedure was consistently superior in producing reliable factor score estimates at both level-1 and level-2 -- rather, findings were contextual. At level-1, scores extracted from multilevel models that disaggregated true level-1 and level-2 scores were generally correlated highest with true scores. At level-2, correlations between true scores and factor scores depended on the underlying level-2 factor structure. With the same number of factors at level-1 and level-2, aggregated and disaggregated approaches performed similarly with 250 level-2 clusters. With only 50 level-2 clusters, aggregated approaches outperformed disaggregated approaches. With fewer factors at level-2 compared to level-1, multilevel measurement models were superior regardless of the number of level-2 clusters. Incorporating measurement non-invariance into measurement models did not substantially improve factor score estimates in the presence of non-invariance at level-2. Simulation results were used to inform a subsequent empirical analysis and demonstration, and were translated into recommendations for research in practice. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]