Statistics textbooks for undergraduates have not caught up with the enormous amount of analysis of Internet data that is taking place these days. Case studies that use Web server log data, Internet survey data or Internet network traffic data are rare in undergraduate Statistics education. This paper summarizes the results of research in three areas of Internet data analysis: users' web browing behavior, user demographics, and network performance. We present some of the main questions analyzed in the literature, some unsolved problems, and some typical data analysis methods used. We illustrate the questions and the methods with large data sets. The data sets were obtained from the publicly available pool of data. Those data sets had to be processed and transformed to make them available for classroom exercises. The processed data sets as well as more material for classes, are available at a web site with address that can be obtained from the main author.

Statistics textbooks for undergraduates have not caught up with the enormous amount of analysis of Internet data that is taking place these days. Case studies that use Web server log data, Internet survey data or Internet network traffic data are rare in undergraduate Statistics education. This paper summarizes the results of research in three areas of Internet data analysis: users' web browsing behavior, user demographics, and network performance. We present some of the main questions analyzed in the literature, some unsolved problems, and some typical data analysis methods used. We illustrate the questions and the methods with large data sets. The data sets were obtained from the publicly available pool of data. Those data sets had to be processed and transformed to make them available for classroom exercises. The processed data sets as well as more material for classes, are available at a web site with address that can be obtained from the main author.

In the thesis, we focused on cluster analysis with longitudinal data. In the Study of Women's Health Across the Nation (SWAN), we are interested in identifying trending groups based on repeated hormone observations, and in the association between hormone trends and women's health during menopause. we proposed a Bayesian semi-parametric model with a Dirichlet Process (DP) that allows for irregular data and model-based clustering of women based on their hormone profiles. We identified distinct developmental trajectories for both E2 and FSH hormones, and discussed relationships between profiles and other factors. Urinary Incontinence (UI) is considered as a specific measure of women's health, and we extended our approach to jointly model UI and hormone outcomes in order to explore the association between UI status and the pattern of hormone changes over the menopausal transition.

In the study of Johne's disease in cattle, we developed a particular finite mixture model for diagnosis based on bivariate longitudinal outcomes. The subjects were clustered into three disease phases: uninfected and two categories of disease.

In addition, we discussed Label Switching issues in Bayesian finite and infinite mixture models, and proposed an ad-hoc method to address them to obtain inferences for the parameters on the component level.

The goal of the research was to develop a short screener that deputies could use in the field to help forecast domestic violence incidents in particular households. Based on information collected from over 600 households, forecasting procedures were developed. They are based on a few simple questions that can be asked at the scene, selected from a larger set of over thirty questions. From these questions alone, correct forecasts of future calls for service can be made about 60% of the time. Future calls involving domestic violence misdemeanors and felonies can be correctly forecast about 50% of the time. The 50% figure is especially important because such calls require a law enforcement response and yet are a relatively small fraction of all domestic violence calls for service.

In this paper, we report on the development of a short screening tool that deputies in the Los Angeles Sheriff?s Department could use in the field to help forecast domestic violence incidents in particular households. The data come from over 500 households to which sheriff?s deputies were dispatched in the fall of 2003. Information on potential predictors was collected at the scene. Outcomes were measured during a three month follow-up. The data were analyzed with modern data mining procedures in which true forecasts were evaluated. A screening instrument was then developed based on a small fraction of the information collected. Making the screening instrument more complicated did not improve forecasting skill. Taking the relative costs of false positives and false negatives into account, the instrument correctly forecasted future calls for service about 60% of the time. Future calls involving domestic violence misdemeanors and felonies were correctly forecast about 50% of the time. The 50% figure is especially important because such calls require a law enforcement response and yet are a relatively small fraction of all domestic violence calls for service.

A number of broader policy implications follow. It is feasible to construct a quick-response, domestic violence screener that is practical to deploy and that can forecast with useful skill. More informed decisions by police officers in the field can follow. Although the same kinds of predictors are likely to be effective in a wide variety of jurisdictions, the particular indicators selected will vary in response to local demographics and the local costs of forecasting errors. It is also feasible to evaluate such quick-response threat assessment tools for their forecasting accuracy. But, the costs of forecasting errors must be taken into account. Also, when the data used to build the forecasting instrument are also used to evaluate its accuracy, inflated estimates of forecasting skill are likely.