Detailed Biography:

As a practicing statistician for over twenty years, the heart of my research interests lies in study design. Most of my applied work has been in health services research, especially in mental health and HIV prevention. Sample design is important for those applications. I have worked on several challenging studies that sampled participants through encounters, such as studies of commercial sex workers as they present on the street, homeless persons as they present on the street and various service locations such as shelters, and HIV infected persons as they present for medical care. Those studies require careful designs and analyses to address the inherent length bias and multiplicity issues (some participants with higher frequency of presentation are more likely to be sampled). Many of those studies also have a multi-level structure, such as patients nested in clinics, thus require careful designs and analyses to account for the multilevel structure. I am interested in a variety of experimental design issues such as the design of multilevel studies (randomization at the clinic level vs. randomization at the individual level vs. split-plot designs that randomize at multiple levels; the trade-off between having more groups vs. having more individuals per group, etc.), the use of factorial designs to enhance our ability to understand the impact of components of multi-faceted interventions, and to help develop more effective and robust interventions, etc. More broadly, my interests in study design also includes working with my collaborators to clarify research questions, so as to design the data collection strategy to address the key research questions effectively. In addition to study design issues, my interests in statistical methods include transformation models, model robustness, causal inference, multilevel models for clustered and/or longitudinal data, nonparametric and semi-parametric methods, etc. Several topics I am exploring currently include the design of longitudinal assessments (how often to follow-up, and when -- not necessarily at equal intervals), analysis for multi-phase screening data (nonresponse weight construction, etc.), model selection for propensity score methods, sensitivity analysis for non-ignorable missing data, automated testing of survey instruments, etc. I am affiliated with UCLA's Neuropsychiatric Institute's Center for Community Health and Health Services Rsearch Center, and the AIDS Institute.