Small Area Estimation and Microsimulation Modeling

Associate Professor Azizur Rahman, PhD, is a statistician and data scientist with expertise in both developing and applying novel methodologies, models and technologies. He is the Leader of “Statistics and Data Mining Research Group” at the Charles Sturt University (CSU), and able to assist in understanding multi-disciplinary research issues within various fields including how to understand the individual activities which occur within very complex scientific, behavioural, socio-economic and ecological systems. His research encompasses issues in simple to multi-facet analyses in various fields ranging from the statistical sciences to the law and legal studies. He has more than 100 scholarly publications including a few books. Prof. Rahman’s research is funded by the Australian Federal and State Governments, and he serves on a range of editorial boards including the International Journal of Microsimulation (IJM) and Sustaining Regions. He obtained several awards including the SOCM Research Excellence Award 2018 and the CSU-RED Achievement Award 2019. Professor Ann Harding, AO, is an Emeritus Professor of Applied Economics and Social Policy at the National Centre for Social and Economic Modelling (NATSEM) of the University of Canberra. She was the founder and inaugural Director of this world class Research Centre for more than sixteen years, and also a co-founder of the International Microsimulation Association (IMA) and served as the inaugural elected president of IMA from 2004 to 2011. She is a fellow of the Academy of the Social Sciences in Australia. She has more than 300 publications including several books in microsimulation modeling.

• Table of ContentsPrefaceIntroductionIntroduction Main Aims of the BookGuide for the ReaderConcluding RemarksSmall Area EstimationIntroductionSmall area estimationAdvantages of small area estimationWhy small area estimation techniques?Applications of small area estimationApproaches to small area estimationDirect estimationHorvitz-Thomposn (H-T) estimatorGeneralized regression (GREG) estimatorModified direct estimatorDesign-based model-assited estimatorsA comparison of direct estimatorsConcluding remarksIndirect Estimation: Statistical ApproachesIntroduction Implicit models approachSynthetic estimatonComposite estimationDemographic estimationComparison of various implicit models based indirect estimationExplicit models approachBasic area level modelBasic unit leve modelGeneral linear mixed modelComparison of various explicit models based indirect estimationMethods for estimating explicit modelsE-BLUP approachEB approachHB approachA comparison of three methodsConcluding remarksIndirect Estimation: Geographic ApproachesIntroductionMicrosimulation modelingProcess of microsimulationTypes of microsimulation modelsAdvantages of microsimulation modelingMethodologies in microsimulation modeling technologyTechniques for creating spatial microdataStatistical data matching or fusionIterative proportional fittingRepeated weighting methodReweightingCombinatorial optimisation reweighing approachThe simulated annealing method in COAn illustration of CO process for hypothetical dataReweighting: The GREGWT approachTheoretical settingHow does GREGWT generate new weights?Explicit numerical solution for a hypothetical dataA comparison between GREGWT and COConcluding remarksBayesian Prediction-Based Microdata SimulationIntroductionThe basic stepsThe Bayesian prediction theoryThe multivariate modelThe prior and posterior distributionsThe linkage modelPrediction for moedling unobserved population unitsConcluding remarksMicrosimulation Modelling Technology for Small Area EstimationIntroductionData sources and issuesThe Census DataSurvey DatasetsSurvey DatasetsMMT based Model SpecificationModel inputsGenerating small area synthetic weightsModel inputs Generating small area synthetic weightsModel inputsGnerating small area synthetic weightsModel outputsHousing stressDefinitionMeasures of housing stressA comparison of various measuresSmall area estimation of housing stressInputs at the second stae modelFinal model outputsConcluding remarksApplications of the MethodologiesIntroduction Results of the model: A general viewModel accuracy reportScenarios of housing stress under various measuresDistribution of housing stress estimationLorenz curve for housing stress estimatesProportional cumulative frequency graph and index of dissimilarityScenarios of households and housing stress by tenuresEstimation of households in housing stress by spatial scalesResults for different statesResults for various statistical divisionsResults for various statistical subdivisionsSmall area estimates: Number of households in housing stressEstimated numbers of overall households in housing stressEstimated numbers of buyerhouseholds in housing stress Estimated numbers of public renter households in housing stressEstimated numbers of private renter households in housing stressEstimated numbers of total renter households in housing stressSmall area estimates: Percentage of households in housing stressPercentage estimates of housing stress for overall householdsPercentage estimates of housing stress for buyer householdsPercentage estimates of housing stress for public renter householdsPercentage estimates of housing stress for private renter householdsPercentage estimates of housing stress for total renter householdsConcluding remarksAnalysis of Small Area Estimates in Capital CitiesIntroduction Scenarios of the results for major capital citiesTrends in housing stress for some major citiesMapping the estimates at SLA levels within major citiesSydneyHousing stress estimates for overall householdsSmall area estimation by household's tenure types MelbourneHousing stress estimates for overall householdsSmall area estimation by household's tenure types BrisbaneHousing stress estimates for overall householdsSmall area estimation by household's tenure typesAdelaideHousing stress estimates for overall householdsSmall area estimation by household's tenure typesCanberraHousing stress estimates for overall householdsSmall area estimation by household's tenure typesHobartHousing stress estimates for overall householdsSmall area estimation by household's tenure typesDarwinHousing stress estimates for overall householdsSmall area estimation by household's tenure typesConcluding remarksValidation and Measure of Statistical ReliabilityIntroductionSome validation methods in the literatureNew approaches to validating housing stress estimationStatistical significance test of the MMT estimatesResults of the statistical significance testAbsolute standardised residual estimate (ASRE) analysisResults from the ASRE analysisMeasure of statistical reliability of the MMT estimatesConfidence interval estimationResults from the estimates of confidence intervalsConcluding remarksConclusions and Computing CodesIntroduction Summary of major findingsLimitationsAreas of further studiesComputing codes and programmingThe general model file codesSAS programming for reweithing algorithmsThe second stage program file codesConcluding remarksAppendices.

“The authors describe theories and provide practical applications in economics, behavioural sciences, health sciences, business, psychology, environmental sciences, transportation problems, urban planning, and computational issues in this book. … Some unique features of this book are the following. The historical quotes from early time as far as fifth century are amusing. The references related to SAE are not only thorough but also up to date. The guidance for the readers does ease the readability. I enjoyed reading this comprehensively written book. I recommend this book to sociologists, economists, geographers, statistics and computing professionals.”—Ramalingam Shanmugam, in the Journal of Statistical Computation and Simulation, June 2019"The book aims at introducing modern statistical small area estimation methodologies into the framework of spatial microsimulation modelling for a comprehensive presentation, providing a novel approach with much potential in comparative social research and regional statistics production. In my opinion, the strongest methodological developments are in the techniques of generating synthetic spatial microdata at small area levels. This book will be attractive for students, in economics, social sciences and statistics in particular. The increasing use of both SAE and microsimulation methods in different areas of society, such as social planning by government institutions and official or public statistics production by national and international statistical agencies. Finally, I want to congratulate the authors for writing a nice and well readable book on a quite complicated topic." —Prof. Risto Lehtonen, University of Helsinki". . .an interesting read for both beginning and more experienced microsimulation modellers. The two authors are well known within the microsimulation community. In this book, they share their experiences and insights into both the more theoretical and empirical aspects of microsimulation modelling. Across disciplines, there are several approaches towards the simulation or projection of small area statistics. However, since these different disciplines make use of different terminologies, there is less cross-pollination than expected (or hoped for). The aim of this book is to show and explain different approaches of small area estimation that are used in different research fields. The book gives an extensive theoretical and empirical overview of different microsimulation techniques and can be of relevance to researchers who want to expand their knowledges on ways to estimate small area characteristics."~International Journal of Microsimulation"The authors begin with a detailed classification tree of small area estimation techniques. The text then proceeds to review and describe these techniques. A familiarity with regression techniques and survey methods is assumed throughout. The text then proceeds to present some new small area estimation techniques, validation methods, and a detailed worked example. The appendices provide further details of the worked example and SAS code for the generalized regression weighting tool (GREGWT) method."~Douglas Dover, International Society for Clinical Biostatistics