This study involves investigating the impacts of ignoring nested data structure in Rasch/1PL item response theory (IRT) model via a two-level and three-level hierarchical generalized linear model (HGLM). Currently, Rasch/IRT models are frequently used in educational and psychometric researches for data obtained from multistage cluster samplings, which are more likely to violate the assumption of independent observations of examinees required by Rasch/IRT models. The violation of the assumption of independent observation, however, is ignored in the current standard practices which apply the standard Rasch/IRT for the large scale testing data. A simulation study (Study Two) was conducted to address this issue of the effects of ignoring nested data structure in Rasch/IRT models under various conditions, following a simulation study (Study One) to compare the performances of three methods, such as Penalized Quasi-Likelihood (PQL), Laplace approximation, and Adaptive Gaussian Quadrature (AGQ), commonly used in HGLM in terms of accuracy and efficiency in estimating parameters. As expected, PQL tended to produce seriously biased item difficulty estimates and ability variance estimates whereas almost unbiased for Laplace or AGQ for both 2-level and 3-level analysis. As for the root mean squared errors (RMSE), three methods performed without substantive differences for item difficulty estimates and ability variance estimates in both 2-level and 3-level analysis, except for level-2 ability variance estimates in 3-level analysis. Generally, Laplace and AGQ performed similarly well in terms of bias and RMSE of parameter estimates; however, Laplace exhibited a much lower convergence rate than that of AGQ in 3-level analyses. The results from AGQ, which produced the most accurate and stable results among three computational methods, demonstrated that the theoretical standard errors (SE), i.e., asymptotic information-based SEs, were underestimated by at most 34% when 2-level analyses were used for the data generated from 3-level model, implying that the Type I error rate would be inflated when the nested data structures are ignored in Rasch/IRT models. The underestimated theoretical standard errors were substantively more severe as the true ability variance increased or the number of students within schools increased regardless of test length or the number of schools. / Ph. D.

Pharmacodynamic processes and disease progression are increasingly characterized with pharmacometric models. However, modelling options for discrete-type responses remain limited, although these response variables are commonly encountered clinical endpoints. Types of data defined as discrete data are generally ordinal, e.g. symptom severity, count, i.e. event frequency, and time-to-event, i.e. event occurrence. Underlying assumptions accompanying discrete data models need investigation and possibly adaptations in order to expand their use. Moreover, because these models are highly non-linear, estimation with linearization-based maximum likelihood methods may be biased. The aim of this thesis was to explore pharmacometric methods and novel models for discrete data through (i) the investigation of benefits of treating discrete data with different modelling approaches, (ii) evaluations of the performance of several estimation methods for discrete models, and (iii) the development of novel models for the handling of complex discrete data recorded during (pre-)clinical studies. A simulation study indicated that approaches such as a truncated Poisson model and a logit-transformed continuous model were adequate for treating ordinal data ranked on a 0-10 scale. Features that handled serial correlation and underdispersion were developed for the models to subsequently fit real pain scores. The performance of nine estimation methods was studied for dose-response continuous models. Other types of serially correlated count models were studied for the analysis of overdispersed data represented by the number of epilepsy seizures per day. For these types of models, the commonly used Laplace estimation method presented a bias, whereas the adaptive Gaussian quadrature method did not. Count models were also compared to repeated time-to-event models when the exact time of gastroesophageal symptom occurrence was known. Two new model structures handling repeated time-to-categorical events, i.e. events with an ordinal severity aspect, were introduced. Laplace and two expectation-maximisation estimation methods were found to be performing well for frequent repeated time-to-event models. In conclusion, this thesis presents approaches, estimation methods, and diagnostics adapted for treating discrete data. Novel models and diagnostics were developed when lacking and applied to biological observations.