Program

The International Conference for Health Statistics in the Pacific Islands brings together statisticians and health researchers from the Pacific Islands and the wider international community. Roughly half of the conference will consist of training from statistical experts, and the other half of the program will consist of seminars on health projects where statistics has been used as a supporting tool. Abstracts for these training sessions and seminars will be added as they become available. A program appears below. Note that there are two streams - one for researchers who are new to the field of statistics, and for more advanced statisticians. Delegates are free to attend either stream as they choose.

Advanced Stream:

Time Tuesday, 5 July
Wednesday, 6 July
Thursday, 7 July
Friday, 8 July

7:30-8:30
Pick up registration material

8:30-9:00 Opening Remarks
Pick up registration Material
Pick up registration material
Pick up registration material

Mark Griffin

9:00-10:30 Mixed Effects and Clustered Data I Mixed Effects and Clustered Data ILinear regression makes the simplifying assumption that each observation is completely independent; however in a large number of situations this assumption is violated. Data can appear in clusters for example when patients are recruited into a study from a number of hospitals; in general different patients from the same hospital will be more similar than patients from different hospitals. Longitudinal studies are a special case of clustered data, where a number of measurements are collected on each person, and measurements within each person are more similar than measurements between people. In this seminar I will introduce the first of two major families of methods for dealing with clustered and longitudinal data, namely mixed effects models. Applied Multivariate Statistical Analysis IIApplied Multivariate Statistical Analysis IIThe second session will cover topics in classification and clustering. In data mining or machine learning, they are called supervised learning and unsupervised learning, respectively. There have been exciting new developments in recent years motivated by high volume/dimensional data from informatics to genomics microarrays/sequencing data. Some of the latest algorithms will be introduced with real world application examples. Introductory Survey SamplingIntroductory Survey SamplingOne clear advantage of random sampling is its ability to generalize data from a sample to a population. The workshop will demonstrate the use of the following methods of sampling through examples: simple random, stratified, clustered, and systematic. Advantages and disadvantages of each method will be discussed and there will be a focus on identifying the appropriate time for using probability sampling in general and each method in particular. Finally, there will be a discussion of sample sizes and confidence intervals from probability samples. Operations Research I - Programming Operations Research I - ProgrammingMathematical Programming is a collection of techniques for finding the optimal value (maximum or minimum) of an objective function of several decision variables satisfying a number of constraints. It is said to be a Linear Programming (LP) problem if it meets the condition of strict linearity in its objective function and constraints. This workshop presents an introductory approach and the application of the technique to practical problems in the industries. With discussions of basic terminologies and characteristics of LP problems and basic of modeling these problems, this workshop wishes to instill in attendees the ability to formulate any LP problem. Specific examples are discussed and written in the standard form. We will consider only two methods; emphasis will be given to the graphical method, which has limitations to solving LP problems that involve only two variables. Existence of different forms of solutions arising out of special cases of constraints is then presented. The four cases that exist are: unique optimal solution, alternative solution, unbounded solution, and infeasibility. The second solution method, namely the Simplex Method, is then introduced. This algorithm for solving LP problems, starts with an initial solution and then evaluates adjacent solutions, moves to solutions that improve the objective function, and continues until an optimum solution is obtained. The strength in this method is its ability to efficiently solve large LP problems.

Mark Griffin
Dongseok Choi
Justin S Fisher
Karuna Reddy

11:00-12:30 Mixed Effects and Clustered Data II Mixed Effects and Clustered Data II In the first seminar of this pair I introduced one family of methods for dealing with clustered and longitudinal data, namely mixed effects models. In this second seminar I will introduce the other major family of methods, generalized estimating equations. Mixed effects models are very popular when the goal of a study is to characterize the different sources of variance within a dataset. Generalised estimating equations are popular when the major goal of a study is to produce a model of the average set of result over the clusters or patients, for instance to produce a model for how the characteristics of the average patient are changing over time. Basic Environmental Epidemiology IBasic Environmental Epidemiology IEnvironmental epidemiology is concerned with the effects of external exposures on disease risks in human populations. This series aims to provide statisticians with an overview of basic methods used and some of the current “hot topics” in the field, illustrated with examples drawn from radiation and air pollution studies. This first lecture will describe study designs (cohort, case-control, and various hybrids) and the statistical workhorses for binary disease outcomes (logistic, Poisson, and Cox regression). This lecture will conclude with some extensions for modeling more complex exposure-time-response relationships. A Variety of Health Statistics ApplicationsA Variety of Health Statistics Applications Methods to Account for Deaths and Missing Data in a Longitudinal Study of Elderly Women
Steven Bowe, David Sibbritt, Patrick Mcelduff, Anne Young
This seminar will discuss methods developed for analysing longitudinal changes in health related quality of life. Three methods for incorporating death into longitudinal studies of elderly populations and another method that attempts to deal with longitudinal missing data which may be missing not at random (MNAR). The methods were applied to the Australian Longitudinal Study on Women’s Health (1996-2005) data with follow-up surveys every three years. A case study examined differences in elderly women with and without diabetes over time and the impact of including deaths and other missing data. Methods suggest ignoring deaths and other missing data can lead to biased results towards study survivors.

Out-of-School Activities of Three Fiji Primary Schools
Jeremy Dorovolomo
A quantitative study, this investigation uses the Statistical Package for the Social Sciences (SPSS) to perform Analysis of Variance (ANOVA) on rural-urban and gender comparisons of after-school activities. This involved sixty-three class-four pupils of three Fiji primary schools. Among findings, rural children tend to ‘play outside’ more than their urban counterparts. However, levels of television viewing are consistently high for both locations, diminishing children’s opportunities to be physically active. The low participation rate in after-school community sports also suggests a need for community clubs and sporting federations to provide structures for positive, inclusive, and developmentally appropriate participation.

Does Greater Exposure to Anti-Smoking Advertising Prevent Smoking Relapse?
Melanie Wakefield, Steven Bowe, Sarah Durkin, Mathew Spittal, Hua-Hie Yong, Ron Borland, and Julie Simpson
The aim of this seminar is to discuss the statistical methods and findings of our attempt to determine whether the timing and intensity of exposure to tobacco control mass media campaigns may prevent relapse among a cohort of smokers who have recently quit. Estimates of population exposure to televised anti-tobacco advertising from commercial media monitoring sources were merged to a cohort study of recently quit adult smokers, using the data collection date and postcode information. Participants were drawn from the Australian arm of the International Tobacco Control Four (ITC-4) Country survey (2002-2008). Uniquely, additional participants were recruited each year to replace those lost to attrition.
Operations Research II - Integer & Nonlinear Programming Operations Research II - Integer & Nonlinear Programming In this workshop we continue the discussion of linear programming, focusing on formulating and solving LP problems. Included is a demonstration of solving linear programming problems using Excel Solver. We then consider extensions that allow for integer restricted decision variables, multiple goals, and nonlinear objective functions and constraints. The session concludes with a group discussion of potential health care problems to which mathematical programming could be applied.

Mark Griffin
Duncan Thomas
Steven Bowe, Jeremy Dorovolomo
James J. Cochran

1:30-3:00 Applied Multivariate Statistical Analysis IApplied Multivariate Statistical Analysis IThis is the first part of two sessions on applied multivariate analysis designed for the advanced stream students. The first session will start with brief reviews of definitions, terminology and multivariate normal distribution. The main subjects will include classical multivariate analysis such as testing for a multivariate mean vector, principal component analysis, factor analysis. All examples will be presented with R, which is a free program for statistical computing and graphs (www.r-project.org) and one of the most popular statistical computing environments. Example codes will be easily adaptable for other applications. Basic Environmental Epidemiology IIBasic Environmental Epidemiology II The second lecture will provide a general overview of current problems in longitudinal analysis, spatio-temporal modeling of exposures and diseases, aggregate data analysis (ecologic correlation), exposure measurement error, and gene-environment interactions. I will conclude with a discussion of some approaches to evaluation of the health benefits of programs aimed at reducing exposures and policy implications for risk assessment and compensation. Effectively Communication with StudentsEffectively Communication with StudentsWe in the statistics community understand that statistics is an inherently interesting, relevant, important, and enjoyable discipline - unfortunately many of our students and clients don’t seem to share this understanding with us! So how do statisticians help students appreciate that statistics is interesting and relevant and important and enjoyable? Professor Cochran discusses several classroom cases and active learning exercises he has developed and regularly uses to accomplish this goal when teaching introductory Statistics courses. Throughout this session Professor Cochran will emphasize his points with live demonstrations and discussions of several interesting and novel active learning exercises and cases. Card tricks, classroom versions of television game shows, and a teaching case with integrated active learning will be featured. Because many of these exercises are easily transferable across topics, instructor/classroom styles, cultures, national borders, institutions, faculties, programs, and class sizes, it is very likely you will walk away from this session with ideas on how to improve your own teaching (indeed, Professor Cochran will be very disappointed if you don't!). Basics of Nonlinear StatisticsBasics of Nonlinear StatisticsResearchers often recognize that nonlinear models usually fit their data well and often in a more parsimonious manner (with far fewer model parameters), and the corresponding model parameters are usually more scientifically meaningful. This course reviews the essentials of linear regression, and subsequently introduces and illustrates generalized linear models (such as logistic regression), Gaussian nonlinear models, and generalized nonlinear models, focusing on applications. Illustrations are given from the domains of bioassay, relative potency and drug or similar compound synergy useful in biomedical and environmental sciences. Caveats are discussed regarding convergence, diagnostics, and the inadequacy of standard (Wald) confidence intervals.

Dongseok Choi
Duncan Thomas
James J. Cochran
Timothy E. O'Brien

3:30-5:00 Seminar 1 - Mathematical Programming in Health CareSeminar 1 - Mathematical Programming in Health CareMathematical Programming (MP) is commonly used to solve decision making problems in the diverse areas such as sociology, social psychology, demography, political science, economics, education, public health, and many others. In this seminar, mathematical programming is efficiently applied to the topic of survey sampling, in particular construction of strata in stratified random sampling. This approach could be used to estimate or forecast various characteristics of a population by using survey data and simultaneously increase the precision of the resulting estimates. There are always limitations with regards to cost, so data that is easily available from past surveys is used to obtain the desired result without having to conduct a new survey. Real health data (micronutrient status of women in Fiji) from a prior survey is used, and its frequency distribution is estimated after a series of testing and analysis via some statistical packages. The problem of determining the optimal strata boundaries (OSB) is then formulated as an MP problem for which the objective is minimization of the estimated population parameter’s variance. The formulated MP problems are multistage decision problems, and so are solved for the OSBs using dynamic programming. Emphasis is given to the actual data analysis, estimation of the distribution, formulation of the MP problem, and its solution procedure. Seminar 2 - Medical Research in Doctors' OfficesSeminar 2 - Medical Research in Doctors' OfficesIt is possible to investigate important medical questions with limited budgets and personnel. We present four examples of “small science” investigations that have been carried out by physicians and nurses in their own practices in Rochester, Minnesota: accuracy of tympanic thermometers; use of eye lubricant in comatose patients; use of Tylenol for pediatric vaccinations; and Epley’s maneuver for benign positional vertigo. We also describe two examples of large studies in Practice Based Research Networks (PBRNs), which pool data from physician offices across the United States. Seminar 3 - Applications in Clinical TrialsSeminar 3 - Applications in Clinical TrialsThis advanced lecture will give an overview of an actual successful clinical trial program for drug development, and for which the speaker was the principal statistician. The lecture will first give an overview to the first large population based, and successful public clinical trials of the Salk Polio Vaccine. The course will present in detail a case study of the CELLCEPT drug development program. CELLCEPT was a novel drug, approved worldwide, and now a standard of care, for daily use by patients who receive a Kidney, Heart or Liver transplant. The course will discuss the various required Phase I, II, and III clinical studies adopted in the drug development program. The course will explain challenges in conducting clinical trials, defining the data collection, setting up databases, dose selection, challenges of day by day monitoring of accumulating data in the clinical studies, choice of endpoints, and miscellaneous operational “day to day” issues, and strategic, regulatory (FDA,EMEA, MHW, DPI etc.) issues and concerns in the approval of a new drug. Plenary and Wrap-upPlenary and Wrap-upThe concluding session of the 2011 International Conference for Health Statistics in the Pacific Islands (ICHSPI) will feature a discussion of organizations that share the vision and support the efforts of ICHSPI faculty and delegates (with emphasis on Statistics Without Borders, http://community.amstat.org/AMSTAT/StatisticsWithoutBorders/). Various initiatives that have been undertaken by these organizations will be highlighted, and the session will conclude with remarks about potential directions for health care research and the future of the ICHSPI.

Karuna Reddy
Peter Wollan
Chris Barker
James J. Cochran

Introductory Stream:

Time Tuesday, 5 July
Wednesday, 6 July
Thursday, 7 July
Friday, 8 July

7:30-8:30 Pick up registration material

8:30-9:00 Opening Remarks
Pick up registration material
Pick up registration material
Pick up registration material

Mark Griffin

9:00-10:30 Role of Statistics Role of Statistics An introduction to statistical concepts. Measurement and repeatability; distributions; statistics of location and spread; types of data. Statistical models, estimation, and inference; assumptions underlying the model, possible effects of incorrect assumptions, bias, and error. Association, correlation, and causation. Specific examples discussed will include capture-recapture estimation of animal population size; convenience samples, random samples, and self-selected samples for questionnaires; and randomized controlled trials to evaluate drug efficacy. Basics of Statistical Inference IBasics of Statistical Inference IIn this workshop we present an overview of interval estimation. We first define basic statistical terms: population, random sample, population parameters (mean and proportion), and statistics. Next we will talk about variability, and demonstrate sampling distribution. We then talk briefly about Normal Distribution and Central Limit Theorem. We define and calculate confidence intervals for mean and proportion, and discuss the interpretations of confidence intervals, confidence level, and margin of error. Finally, we discuss the sample size needed to obtain the desired precision of estimation. Introduction to Linear Regression IIIntroduction to Linear Regression IIIn this session we will build on the descriptive treatment of regression in part I by introducing the concepts of inference, statistical significance, and confidence intervals. Some attention will be given to the basics of sampling and the assumptions underlying inference in general and regression in particular. We will introduce methods of valuating regression models and point out their limitations as well as their power, touching on choice of variables, outliers, and predictive as well as descriptive aspects of regression. The major focus will be on applications of the techniques of linear regression to real life situations. Fundamentals of Clinical TrialsFundamentals of Clinical TrialsThis introductory course will give an overview to the use of clinical trials in pharmaceutical drug development. The course will define and explain the role of Phase I, II III and IV clinical trials in the U.S drug approval process for the Food and Drug Administration, as well as Global agencies, EMEA, MHK etc. The course will give an overview of the relevant design of experiments, including protocol development, data collection, case report form, data monitoring, database development, specification of hypotheses, and identification of patients/subjects to enroll. The lecture will briefly touch on role of toxicology, carcinogenicity studies, formulation, and chemistry, manufacturing and controls “CMC”. Topics discussed will also include concepts central to designing, conducting and understanding of clinical trials. This includes understanding potential biases, randomization, blinding, equipoise, the role of informed consent, and the role of the Declaration of Helsinki and the rights of experimental subjects. The critical and essential role of statistics in clinical trials is outlined, in terms of specification of statistical hypotheses, placebo control “adequate and well controlled clinical trials”, selecting and planning statistical analyses, data collection, and foremost, understanding and interpreting the results of clinical trials.

Peter Wollan Grazyna Badowski
Mary Gray
Chris Barker

11:00-12:30 Handling Data with Spreadsheets Handling Data with Spreadsheets Quality information depends fundamentally on entering, checking, managing and exploring data well. Spreadsheets are commonly used for entry, and I will show how they can be used effectively. Guidelines to be discussed include:
enter each observation in a separate row;
use a column to represents each variable with a short name in a single top row;
no blank rows or columns;
use validation tools at both entry and checking stages;
include essential data about the data, including where, when, and by whom;
do summaries on separate sheets using pivot tools; and
use filters and graphs to explore. Basics of Statistical Inference IIBasics of Statistical Inference IIIn this workshop we discuss steps in testing hypothesis: statement of null and alternative hypothesis, selection of significance level, calculating test statistics and p-values, making decisions, and interpreting results. We then talk about Type I and II errors and factors affecting them. We look at many scenarios and discuss possible “costs” for each type of error. Then we discuss the power of the test and it’s relation to the two types or errors and sample size. Finally, the difference between statistical and practical significance will be discussed. Beginning of Design of Experiments IBeginning of Design of Experiments IOften, introductory training in biostatistics focuses on methods to analyze data from health-related research studies. However, good study design is as essential, if not more, to draw meaningful conclusions from data. Thus, the purpose of this seminar is to introduce experimental design concepts for biomedical and public health sciences. Several designs will be covered, including parallel groups, repeated measures, crossover designs and others. In each case, we will highlight the importance of aligning study objectives, design and analysis. The strengths of randomized designs will be discussed. Practical advice will be given regarding how to approach design of new studies. Attendees are encouraged to bring examples of studies they have been involved with for discussion. Basics of Panel DataBasics of Panel DataMany health-related investigations follow up individuals over time and observe them at multiple time points. For example, babies may be followed up in studies of their growths. Patients may be followed up for assessing changes in their health by treatment. These follow-up studies lead to “panel data” or “longitudinal data” on individuals who are followed up over time. In this session, we will discuss these “panel data”, focusing on their pros and cons, using multiple examples of health-related investigations. Participants are encouraged to ask questions and take part in discussions.

Ian Westbrooke
Lynne Wilkens
Jodi Lapidus
Yutaka Yasui

1:30-3:00 Graphs & Displays of Data Graphs & Displays of DataGraphs are essential tools for exploring, analyzing, and communicating information from data. The aim is to produce effective graphs, and to avoid falling into traps that common software seems to encourage. I will explain some results about human perception, and review a variety of graph formats. I will suggest how to improve visual impact, including reducing distraction from non-essential elements. I will argue against the use of pie charts and for using 3-dimensional graphs only in interactive contexts; propose less emphasis on bar charts, and recommend dot and scatter and box plots, plus multi-panel graphs for presenting complex data. Introduction to Linear Regression I Introduction to Linear Regression IWe will begin by discussing the acquisition and organization of data, including attention to various types of variables and forms of display. Simple descriptive statistics will be introduced such as mean, median, and standard deviation, with class participation in simple applications. Next will be a brief introduction to geometric concepts, focusing on straight lines. Then the basic ideas of linear regression and correlation will be explored, with applications to everyday examples of relevance to the participants. Attention will be focused on using data to model and interpret relationships. Seminar 3A - Conducting Public Health Research in Collaboration with Native Communities in the U.S.Seminar 3A - Conducting Public Health Research in Collaboration with Native Communities in the U.S.Certain native populations in the U.S. - American Indians, Alaska Native, Native Hawaiians - experience disparities in many acute and chronic health conditions relative to the entire U.S. population. However, there are organizations dedicated to helping native communities find ways to improve their health status. The instructor will share her experience collaborating with native communities to conduct various public health-related research projects. We will introduce the concept of the community-based participatory research model, and discuss some of the viable study designs, data collection and analysis methods used in successful projects. In particular, we will focus on the group-randomized design which is appropriate for community level interventions. Introduction to Logistic RegressionIntroduction to Logistic RegressionLogistic regression is a technique used when the outcome of interest has one of two possibilities, for instance, received vaccine or did not receive vaccine, and you would like to explain, describe, and analyze this outcome variable in relation to multiple “explanatory” variables, e.g., gender of children, residence, and socioeconomic status. Contrast will be made to other analytical methods that conference participants have been introduced to during the conference. This will be a very interactive session and examples will be taken directly from workshop participants, if possible. Participants will be encouraged to share their stories and ask questions.

Ian Westbrook
Mary Gray
Jodi Lapidus
Marcy Winget

3:30-5:00 Seminar 1 - Haiti and East Timor "on the ground" Seminar 1 - Haiti and East Timor "on the ground" Surveys are an important source of demographic, health, and economic information and are used to complement data from civil registries and censuses. Collecting survey data in developing countries requires flexibility in order to address important considerations, such as the mode of collection, sample design strategies, and techniques for minimizing non-sampling errors such as coverage, measurement, and non-response. Flexibility is also needed to address unexpected challenges that researchers often encounter. Participants will also be introduced to ethical and survey research guidelines developed by the United Nations. Seminar 2 - Examples of Analyses using Guam Cancer Data Seminar 2 - Examples of Analyses using Guam Cancer DataIn this workshop we will use examples from the Guam Cancer Registry data to illustrate statistical analysis principals. We will compare cancer rates for different ethnicities on Guam and use 2009 Behavioral Risk Factor Surveillance System (BRFSS) data to compare prevalence of risk factors for these cancers by ethnic group. In a second example, we will investigate a possible temporal relationship between polychlorinated biphenyl (PCB) contamination of the Merizo Lagoon on Guam and cancer deaths in the adjoining village of Merizo. We will use historical data to test whether there was a significant difference in the proportion of deaths due to cancer in Merizo compared with the rest of Guam. Seminar 3B - Practical Examples from health services research (with emphasis on evaluation)Seminar 3B - Practical Examples from health services research (with emphasis on evaluation)This seminar will highlight one or two health services research projects that have been conducted recently in Canada under the theme of evaluation of access to quality and timely cancer care. The analyses and results will be presented although the focus of the seminar will be on the context of the healthcare situation, development of research question(s), types of data available, sharing findings with stakeholders, and lessons learned. Participants will be encouraged to ask questions and share their stories in order to make the seminar as interactive as possible. Plenary and Wrap-upPlenary and Wrap-upThe concluding session of the 2011 International Conference for Health Statistics in the Pacific Islands (ICHSPI) will feature a discussion of organizations that share the vision and support the efforts of ICHSPI faculty and delegates (with emphasis on Statistics Without Borders, http://community.amstat.org/AMSTAT/StatisticsWithoutBorders/). Various initiatives that have been undertaken by these organizations will be highlighted, and the session will conclude with remarks about potential directions for health care research and the future of the ICHSPI.

Justin S Fisher
Lynne Wilkens & Grazyna Badowski
Marcy Winget & Yutaka Yasui
James J. Cochran

Time	Tuesday, 5 July	Wednesday, 6 July	Thursday, 7 July	Friday, 8 July
7:30-8:30	Pick up registration material
8:30-9:00	Opening Remarks	Pick up registration Material	Pick up registration material	Pick up registration material
	Mark Griffin
9:00-10:30	Mixed Effects and Clustered Data I Mixed Effects and Clustered Data ILinear regression makes the simplifying assumption that each observation is completely independent; however in a large number of situations this assumption is violated. Data can appear in clusters for example when patients are recruited into a study from a number of hospitals; in general different patients from the same hospital will be more similar than patients from different hospitals. Longitudinal studies are a special case of clustered data, where a number of measurements are collected on each person, and measurements within each person are more similar than measurements between people. In this seminar I will introduce the first of two major families of methods for dealing with clustered and longitudinal data, namely mixed effects models.	Applied Multivariate Statistical Analysis IIApplied Multivariate Statistical Analysis IIThe second session will cover topics in classification and clustering. In data mining or machine learning, they are called supervised learning and unsupervised learning, respectively. There have been exciting new developments in recent years motivated by high volume/dimensional data from informatics to genomics microarrays/sequencing data. Some of the latest algorithms will be introduced with real world application examples.	Introductory Survey SamplingIntroductory Survey SamplingOne clear advantage of random sampling is its ability to generalize data from a sample to a population. The workshop will demonstrate the use of the following methods of sampling through examples: simple random, stratified, clustered, and systematic. Advantages and disadvantages of each method will be discussed and there will be a focus on identifying the appropriate time for using probability sampling in general and each method in particular. Finally, there will be a discussion of sample sizes and confidence intervals from probability samples.	Operations Research I - Programming Operations Research I - ProgrammingMathematical Programming is a collection of techniques for finding the optimal value (maximum or minimum) of an objective function of several decision variables satisfying a number of constraints. It is said to be a Linear Programming (LP) problem if it meets the condition of strict linearity in its objective function and constraints. This workshop presents an introductory approach and the application of the technique to practical problems in the industries. With discussions of basic terminologies and characteristics of LP problems and basic of modeling these problems, this workshop wishes to instill in attendees the ability to formulate any LP problem. Specific examples are discussed and written in the standard form. We will consider only two methods; emphasis will be given to the graphical method, which has limitations to solving LP problems that involve only two variables. Existence of different forms of solutions arising out of special cases of constraints is then presented. The four cases that exist are: unique optimal solution, alternative solution, unbounded solution, and infeasibility. The second solution method, namely the Simplex Method, is then introduced. This algorithm for solving LP problems, starts with an initial solution and then evaluates adjacent solutions, moves to solutions that improve the objective function, and continues until an optimum solution is obtained. The strength in this method is its ability to efficiently solve large LP problems.
	Mark Griffin	Dongseok Choi	Justin S Fisher	Karuna Reddy
11:00-12:30	Mixed Effects and Clustered Data II Mixed Effects and Clustered Data II In the first seminar of this pair I introduced one family of methods for dealing with clustered and longitudinal data, namely mixed effects models. In this second seminar I will introduce the other major family of methods, generalized estimating equations. Mixed effects models are very popular when the goal of a study is to characterize the different sources of variance within a dataset. Generalised estimating equations are popular when the major goal of a study is to produce a model of the average set of result over the clusters or patients, for instance to produce a model for how the characteristics of the average patient are changing over time.	Basic Environmental Epidemiology IBasic Environmental Epidemiology IEnvironmental epidemiology is concerned with the effects of external exposures on disease risks in human populations. This series aims to provide statisticians with an overview of basic methods used and some of the current “hot topics” in the field, illustrated with examples drawn from radiation and air pollution studies. This first lecture will describe study designs (cohort, case-control, and various hybrids) and the statistical workhorses for binary disease outcomes (logistic, Poisson, and Cox regression). This lecture will conclude with some extensions for modeling more complex exposure-time-response relationships.	A Variety of Health Statistics ApplicationsA Variety of Health Statistics Applications Methods to Account for Deaths and Missing Data in a Longitudinal Study of Elderly Women Steven Bowe, David Sibbritt, Patrick Mcelduff, Anne Young This seminar will discuss methods developed for analysing longitudinal changes in health related quality of life. Three methods for incorporating death into longitudinal studies of elderly populations and another method that attempts to deal with longitudinal missing data which may be missing not at random (MNAR). The methods were applied to the Australian Longitudinal Study on Women’s Health (1996-2005) data with follow-up surveys every three years. A case study examined differences in elderly women with and without diabetes over time and the impact of including deaths and other missing data. Methods suggest ignoring deaths and other missing data can lead to biased results towards study survivors. Out-of-School Activities of Three Fiji Primary Schools Jeremy Dorovolomo A quantitative study, this investigation uses the Statistical Package for the Social Sciences (SPSS) to perform Analysis of Variance (ANOVA) on rural-urban and gender comparisons of after-school activities. This involved sixty-three class-four pupils of three Fiji primary schools. Among findings, rural children tend to ‘play outside’ more than their urban counterparts. However, levels of television viewing are consistently high for both locations, diminishing children’s opportunities to be physically active. The low participation rate in after-school community sports also suggests a need for community clubs and sporting federations to provide structures for positive, inclusive, and developmentally appropriate participation. Does Greater Exposure to Anti-Smoking Advertising Prevent Smoking Relapse? Melanie Wakefield, Steven Bowe, Sarah Durkin, Mathew Spittal, Hua-Hie Yong, Ron Borland, and Julie Simpson The aim of this seminar is to discuss the statistical methods and findings of our attempt to determine whether the timing and intensity of exposure to tobacco control mass media campaigns may prevent relapse among a cohort of smokers who have recently quit. Estimates of population exposure to televised anti-tobacco advertising from commercial media monitoring sources were merged to a cohort study of recently quit adult smokers, using the data collection date and postcode information. Participants were drawn from the Australian arm of the International Tobacco Control Four (ITC-4) Country survey (2002-2008). Uniquely, additional participants were recruited each year to replace those lost to attrition.	Operations Research II - Integer & Nonlinear Programming Operations Research II - Integer & Nonlinear Programming In this workshop we continue the discussion of linear programming, focusing on formulating and solving LP problems. Included is a demonstration of solving linear programming problems using Excel Solver. We then consider extensions that allow for integer restricted decision variables, multiple goals, and nonlinear objective functions and constraints. The session concludes with a group discussion of potential health care problems to which mathematical programming could be applied.
	Mark Griffin	Duncan Thomas	Steven Bowe, Jeremy Dorovolomo	James J. Cochran
1:30-3:00	Applied Multivariate Statistical Analysis IApplied Multivariate Statistical Analysis IThis is the first part of two sessions on applied multivariate analysis designed for the advanced stream students. The first session will start with brief reviews of definitions, terminology and multivariate normal distribution. The main subjects will include classical multivariate analysis such as testing for a multivariate mean vector, principal component analysis, factor analysis. All examples will be presented with R, which is a free program for statistical computing and graphs (www.r-project.org) and one of the most popular statistical computing environments. Example codes will be easily adaptable for other applications.	Basic Environmental Epidemiology IIBasic Environmental Epidemiology II The second lecture will provide a general overview of current problems in longitudinal analysis, spatio-temporal modeling of exposures and diseases, aggregate data analysis (ecologic correlation), exposure measurement error, and gene-environment interactions. I will conclude with a discussion of some approaches to evaluation of the health benefits of programs aimed at reducing exposures and policy implications for risk assessment and compensation.	Effectively Communication with StudentsEffectively Communication with StudentsWe in the statistics community understand that statistics is an inherently interesting, relevant, important, and enjoyable discipline - unfortunately many of our students and clients don’t seem to share this understanding with us! So how do statisticians help students appreciate that statistics is interesting and relevant and important and enjoyable? Professor Cochran discusses several classroom cases and active learning exercises he has developed and regularly uses to accomplish this goal when teaching introductory Statistics courses. Throughout this session Professor Cochran will emphasize his points with live demonstrations and discussions of several interesting and novel active learning exercises and cases. Card tricks, classroom versions of television game shows, and a teaching case with integrated active learning will be featured. Because many of these exercises are easily transferable across topics, instructor/classroom styles, cultures, national borders, institutions, faculties, programs, and class sizes, it is very likely you will walk away from this session with ideas on how to improve your own teaching (indeed, Professor Cochran will be very disappointed if you don't!).	Basics of Nonlinear StatisticsBasics of Nonlinear StatisticsResearchers often recognize that nonlinear models usually fit their data well and often in a more parsimonious manner (with far fewer model parameters), and the corresponding model parameters are usually more scientifically meaningful. This course reviews the essentials of linear regression, and subsequently introduces and illustrates generalized linear models (such as logistic regression), Gaussian nonlinear models, and generalized nonlinear models, focusing on applications. Illustrations are given from the domains of bioassay, relative potency and drug or similar compound synergy useful in biomedical and environmental sciences. Caveats are discussed regarding convergence, diagnostics, and the inadequacy of standard (Wald) confidence intervals.
	Dongseok Choi	Duncan Thomas	James J. Cochran	Timothy E. O'Brien
3:30-5:00	Seminar 1 - Mathematical Programming in Health CareSeminar 1 - Mathematical Programming in Health CareMathematical Programming (MP) is commonly used to solve decision making problems in the diverse areas such as sociology, social psychology, demography, political science, economics, education, public health, and many others. In this seminar, mathematical programming is efficiently applied to the topic of survey sampling, in particular construction of strata in stratified random sampling. This approach could be used to estimate or forecast various characteristics of a population by using survey data and simultaneously increase the precision of the resulting estimates. There are always limitations with regards to cost, so data that is easily available from past surveys is used to obtain the desired result without having to conduct a new survey. Real health data (micronutrient status of women in Fiji) from a prior survey is used, and its frequency distribution is estimated after a series of testing and analysis via some statistical packages. The problem of determining the optimal strata boundaries (OSB) is then formulated as an MP problem for which the objective is minimization of the estimated population parameter’s variance. The formulated MP problems are multistage decision problems, and so are solved for the OSBs using dynamic programming. Emphasis is given to the actual data analysis, estimation of the distribution, formulation of the MP problem, and its solution procedure.	Seminar 2 - Medical Research in Doctors' OfficesSeminar 2 - Medical Research in Doctors' OfficesIt is possible to investigate important medical questions with limited budgets and personnel. We present four examples of “small science” investigations that have been carried out by physicians and nurses in their own practices in Rochester, Minnesota: accuracy of tympanic thermometers; use of eye lubricant in comatose patients; use of Tylenol for pediatric vaccinations; and Epley’s maneuver for benign positional vertigo. We also describe two examples of large studies in Practice Based Research Networks (PBRNs), which pool data from physician offices across the United States.	Seminar 3 - Applications in Clinical TrialsSeminar 3 - Applications in Clinical TrialsThis advanced lecture will give an overview of an actual successful clinical trial program for drug development, and for which the speaker was the principal statistician. The lecture will first give an overview to the first large population based, and successful public clinical trials of the Salk Polio Vaccine. The course will present in detail a case study of the CELLCEPT drug development program. CELLCEPT was a novel drug, approved worldwide, and now a standard of care, for daily use by patients who receive a Kidney, Heart or Liver transplant. The course will discuss the various required Phase I, II, and III clinical studies adopted in the drug development program. The course will explain challenges in conducting clinical trials, defining the data collection, setting up databases, dose selection, challenges of day by day monitoring of accumulating data in the clinical studies, choice of endpoints, and miscellaneous operational “day to day” issues, and strategic, regulatory (FDA,EMEA, MHW, DPI etc.) issues and concerns in the approval of a new drug.	Plenary and Wrap-upPlenary and Wrap-upThe concluding session of the 2011 International Conference for Health Statistics in the Pacific Islands (ICHSPI) will feature a discussion of organizations that share the vision and support the efforts of ICHSPI faculty and delegates (with emphasis on Statistics Without Borders, http://community.amstat.org/AMSTAT/StatisticsWithoutBorders/). Various initiatives that have been undertaken by these organizations will be highlighted, and the session will conclude with remarks about potential directions for health care research and the future of the ICHSPI.
	Karuna Reddy	Peter Wollan	Chris Barker	James J. Cochran