(A) comparison of spatio-temporal smoothing methods for small area estimation in complex survey data

Jung, Heesun Graduate School, Yonsei University 2019 국내석사

Sample surveys are widely used to obtain data representative of the population, because it is economically infeasible to obtain detailed data about all of the individuals in a nation. However, a significant problem with survey data is that it is difficult to produce reliable estimates for small areas, because of insufficient data. Therefore many spatial smoothing methods have been suggested to produce reliable estimates of the characteristics of areas with small sample sizes. Also, as healthcare surveys are usually conducted over time, it is important to use accumulated data to show time trends or patterns over years, while producing reliable estimators. In this paper, we report on our work in extending a spatial smoothing method for small area estimation (SAE) in survey data to a space-time situation, and we compare SAE methods using simulation studies and the application of real data. SAE methods can be classified into two main categories: model-based methods and design-based methods. We used the Horvitz-Thompson estimator (Horvitz and Thompson, 1952), the most widely used design-based method, as the design-based method. For the model-based method, five model-based methods described by Mercer (2014) (Naive binomial, Logit-normal, Arcsine, Pseudo-likelihood approach, Effective sample size) were used. In simulation study, we compared SAE methods considering various population, prevalence and sampling scenarios. All model-based methods provided estimates close to true prevalence than HT estimator in space-time situation. Also, estimates tended to be unstable when spatial correlation exist and population has a higher prevalence, regardless of SAE methods and time trends. For the three different models of temporal random effects, there was only a small difference between models. A design-based method using sample weights is the most widely used method for the survey data. However, as for small area estimation, it is more appropriate to use a smoothing method which uses adjacent spatial information than a design-based method in terms of the accuracy of the estimates. In this paper, we showed that a spatial smoothing method can be extended to a situation with space-time structure reflecting time trends and showing patterns over years, while producing a reliable estimator.

Achieving External Validity in Small Domain Estimation and Causal Inference

Li, Katherine University of Michigan ProQuest Dissertations & Th 2024 해외박사(DDOD)

The concept of external validity is to make accurate estimates of target population quantities using a sample of the population. To conduct correct statistical inference, we must also obtain accurate standard error (SE) estimates that reflect the true sampling variation of the estimator. Achieving external validity is complicated by complex sampling designs. This thesis focuses on two specific external validity problems: small domain estimation (SDE) of substantive outcome measures---either of a certain sociodemographic subgroup or geographic area---and causal inference of treatment or exposure effects. The literature often assumes that the analyzed sample is representative of the target population and ignores data collection. Without either randomization of treatment or of selection, biases can be induced to SDE or causal effects. However, accounting for complex sampling design features is a challenge: to obtain unbiased estimates, the model must be correctly specified and incorporate design variables including survey weights. This thesis develops statistical methods to improve the external validity of SDE and causal inference by integrating multiple data sources and leveraging sampling design characteristics. In Chapter 2, we extend the multilevel regression and poststratification (MRP) framework such that we can obtain inferences for subdomains that are partially defined by variables that are available in the sample only. MRP stabilizes small domain estimates by fitting multilevel models and adjusts for selection bias by poststratifying on auxiliary variables, which are population characteristics predictive of the analytic outcome. However, its use is limited by the availability of the joint distribution of the auxiliary variables. By embedding an additional step that estimates the full poststratifier joint distribution, we can correct for the bias that is incurred by omitting the incomplete poststratifying variable from the classical MRP estimation procedure. In Chapter 3, we compare two methods for obtaining small area prevalence estimates (SAE) of a binary outcome when the sample has complex sampling design characteristics (weights, clusters, and strata). The SAEs are typically obtained via predictions from generalized linear mixed models (GLMMs) to increase estimation precision. However, obtaining the SAE variance is tricky as there is no closed-form equation for estimation, aspects that are further complicated by a complex sampling design. We compare two methods for SAE: using a weighted GLMM with jackknife replication to account for sampling design, and using an unweighted GLMM with the weighted finite population Bayesian bootstrap (WFPBB) to account for the sampling design. In Chapter 4, we obtain the population average treatment effect (PATE) from observational studies in contexts where the study sample is a subset of the target population. We focus on comparing two methods: Augmented Inverse Probability Weighting (AIPW) and Penalized Spline of Propensity Methods for Treatment Comparison (PENCOMP). These are both “doubly-robust” procedures in that consistent PATE can be obtained as long as either the treatment or outcome models are correctly specified. However, these methods lose their double-robustness property when the sample is not a simple random sample and the selection mechanism is associated with the treatment mechanism. We show that PENCOMP can be doubly-robust in this case as long as the selection weights are incorporated in the outcome model, and is considerably more efficient than equivalent AIPW methods that also account for complex sample designs.

Uncertainty Quantification for Model Selection, Rate Estimation and Related Topics

Li, Yuanyuan University of California, Davis ProQuest Dissertat 2022 해외박사(DDOD)

Uncertainty measure is used to quantify to what extent a conclusion is unsure, adding critical transparencyfor the safe deployment and use of data-driven decisions. In this dissertation, we propose uncertaintyquantification methods in three different areas. The first topic is developing measures of uncertaintyfor a model selection procedure. The second topic concerns bias correction and variance estimation in inferenceabout rates with sampling errors in the denominators. The third topic is obtaining statistical evidenceof bias in election polls based on a post-election data analysis. All projects provide real-data examples fordemonstrating their applications.

Priority Setting for Achieving Universal Health Coverage in Nigeria: A Spatial and Temporal Analysis and Cost-Benefit Analysis

Kawakatsu, Yoshito ProQuest Dissertations & Theses University of Wash 2023 해외박사(DDOD)

Universal Health Coverage (UHC) is an urgent global priority outlined in the Sustainable Development Goals (SDGs) to ensure the accessibility of health services for all people without causing financial hardship. If current progress continues to 2030, 37% to 61% of the global population will not be covered by essential health services. Therefore, we need to accelerate the increase of service coverage to achieve the UHC target by 2030.There are three specific aims of this dissertation; 1) To identify both individual and contextual factors that are consistently associated with utilization of nine essential maternal and child health services (i.e., ANC, facility-based delivery, modern contraceptive use, immunizations, and childhood illnesses), across survey years and household geolocations, using five national representative cross-sectional surveys in Nigeria; 2) To estimate grid-level coverage of selected essential MCH services in Nigeria using generalized additive models (GAMs) and Gradient Boosting (GB) 3) To estimate required costs and avoidable child deaths by increasing selected essential health service coverage in each community, and to identify the priority sub-national areas.This dissertation emphasizes the importance of multi-dimensional priority setting in achieving Universal Health Coverage in Nigeria. By identifying the factors influencing health service utilization, assessing regional disparities, estimating required costs, and quantifying potential impacts, policymakers can make evidence-based decisions to maximize the efficiency and effectiveness of healthcare interventions. The findings and recommendations of this research contribute to the broader global agenda of achieving UHC and improving health outcomes for all populations, particularly in low- and middle-income countries.

Bootstrapping Mean Squared Error in Model-Based Small Area Estimation

손정현 부산대학교 대학원 2008 국내석사

소지역들로 구성되는 전체 모집단의 표본조사 결과를 바탕으로 소지역들의 관심 있는 특성값을 추정하게 되면 대체로 소지역의 표본크기가 매우 작아서 추정의 신뢰성이 떨어진다. 이런 문제점을 해결하기 위해 소지역들의 보조정보를 공변량으로 하는 모형기반 추정법을 이용하며, 특히, 소지역들 간의 변동을 모형에 반영한 변량효과모형이 자주 이용된다. 선형혼합모형(LMM)을 특별한 경우로 포함하는 일반화선형혼합모형(GLMM)이 사용되는데 본 연구에서는 목표변수가 이항반응인 경우에 로지스틱 혼합모형을 가정하고 소지역 비율을 추정한다. 추출된 표본 데이터와 모형에서 예측한 값을 합성하여 얻어지는 소지역 비율 추정량에 대한 평균제곱오차(MSE)의 추정에 중점을 둔다. 모수적 붓스트랩 근사방법을 제안하고 선형근사에 근거한 추정법과 비교하였다. 변량효과의 정규성 가정에 대한 추정방법들의 민감성을 파악하기 위해 몇 가지 분포함수에 따라 모의실험을 수행한 결과 붓스트랩 근사방법이 선형근사 추정에 비해 변량효과의 분포함수에 관계없이 좀 더 정확한 결과를 나타내었다.

소지역 추정량 실증연구

김규영 忠南大學校 大學院 2002 국내석사

Small area estimation for analysis of relative-risk using HGLM

이명지 서울대학교 대학원 2004 국내석사

소지역 추정을 위한 표본설계

홍태경 成均館大學校 大學院 2002 국내석사

Population Estimation of Small Urban Area by Using Remote Sensing Image

Jia Liu 부산대학교 2014 국내석사

현재 존재한 인구 추정 방법에 바탕으로 이 논문은 도시 인구의 추정 방법에 대하여 특성과 부족을 분석하고 토지 이용 밀도 방법을 향상시키다. 이 연구에서 중국 공하지역 있는 안산시 철서구를 대상으로 지정하고 이 연구에서는 원격 탐사 이미지에서 생활 영역을 통해 도시 인구의 생활 유형과 인구 사이의 수학적 관계를 설정한다. 즉, 이 공하지역의 인구를 추정하기 위해서 개선된 토지 이용 밀도 모델을 사용한다. 본 논문에서 관련된 두 인구 추정방법이 있다. 하나는 토지 이용 밀도 방법이고 나머지 볼륨 방법이다. 존재한 토지 이용 밀도 방법을 불륨 방법을 보다 잘 적합니다. 따라서 우리는 토지 이용 밀도 방법을 기초하여 개선된 밀도 방법을 만든다. 이 방법은 직접 도시 인구 밀도를 추정하는 수학 모델을 통해서 무작위로 선택된 인구 밀도와 인구 밀도가 작업 부하를 감소시키기 위하여 추정 요구하지 않는 인구 통계데이터를 사용한다. 이것은 거주 지역 없는 영역의 효과를 고려하고 있다. 관련된 모든 인구 추정 방법의 결과의 비교를 통해서 개선 된 토지 이용 강도 방법이 오류가 가장 작은 수 있기 때문에 이 방법이 가장 좋은 방법이다.