UCLA

In health and educational research, planned-missing-data designs have been used to reduce the number of variables collected on participants, thus reducing respondent burden and the number of resources necessary for study. The purpose of this dissertation research is to develop and improve analysis strategies for planned-missing-data designs, with specific applications to partial mouth recording protocols in oral health studies and balanced incomplete block designs in large-scale educational survey assessments. For the oral-health examination, multidimensional item response theory models (MIRT) are investigated in addition to multiple imputation strategies from hierarchical normal models to recover information on periodontal disease status when data are collected on only half of the mouth. Using data from the National Assessment of Educational Progress (NAEP), complex MIRT models are investigated to improve the estimation of population ability characteristics as well as to explore the potential for other components of academic to be measured from the same data.