Nonresponse Mechanism Research Articles

HIV estimation using population-based surveys with non-response: A partial identification approach.

HIV estimation using data from the demographic and health surveys (DHS) is limited by the presence of non-response and test refusals. Conventional adjustments such as imputation require the data to be missing at random. Methods that use instrumental variables allow the possibility that prevalence is different between the respondents and non-respondents, but their performance depends critically on the validity of the instrument. Using Manski's partial identification approach, we form instrumental variable bounds for HIV prevalence from a pool of candidate instruments. Our method does not require all candidate instruments to be valid. We use a simulation study to evaluate and compare our method against its competitors. We illustrate the proposed method using DHS data from Zambia, Malawi and Kenya. Our simulations show that imputation leads to seriously biased results even under mild violations of non-random missingness. Using worst case identification bounds that do not make assumptions about the non-response mechanism is robust but not informative. By taking the union of instrumental variable bounds balances informativeness of the bounds and robustness to inclusion of some invalid instruments. Non-response and refusals are ubiquitous in population based HIV data such as those collected under the DHS. Partial identification bounds provide a robust solution to HIV prevalence estimation without strong assumptions. Union bounds are significantly more informative than the worst case bounds without sacrificing credibility.

Transformed Regression Type Estimators in the Presence of Missing Observations: Case Studies on COVID-19 Incidence in Chiang Mai, Thailand

The transformation technique can be used to modify the shape of the variable to improve the performance of the population mean estimator. In the presence of missing data, before estimating the population mean using standard statistical methods, missing data has to be taken care of. In this study, we focus on new transformed regression type estimators when missing data are present in the study variable under the uniform nonresponse mechanism and assume that the population mean of the auxiliary variable is unavailable which usually occurs in practice. An auxiliary variable can assist by increasing the efficacy of estimating the population mean. The bias and mean square error are investigated up to the first order degree approximation using the Taylor series. A simulation and case studies on COVID-19 incidence in Chiang Mai, Thailand are used to assess the performance of the new transformed estimators. The estimated number of COVID-19 patients who have pneumonia and require high-flow oxygen and the estimated daily confirmed cases of COVID-19 in Chiang Mai from the best proposed estimator are around 17 cases and 118 cases, respectively.

Sampling, nonresponse, and quality in travel surveys

Response rates have been continuously declining in most countries for many decades. Therefore, the risk of selectivity and of response bias are steadily increasing. During the workshop, we discussed the use of different lists based on sampling frames, as this can greatly improve the efficiency of data collection. However, the impact of the weighting procedure must be considered. The study of nonresponse mechanism is fundamental for representativeness of the sample of respondents. Furthermore, soft refusal is not limited to not reporting any trips at all, but can also be defined as underreporting trips. We found that it is necessary to improve surveys with the help of technologies or usage logs in order to deal with burdensome and underreporting. At the end of the workshop we came up with research priorities and recommendations for the next conference.

A new imputation method for population mean in the presence of missing data based on a transformed variable with applications to air pollution data in Chiang Mai, Thailand

Introduction: Chiang Mai’s air pollution has risen to number one in the world for the highest level of fine particulate matter which further exacerbates the damage to human health. Fine particulate matter can enter the human body and blood circulation, destroying organ systems, increasing the risk for chronic disease and cancer, despite not having smoking habits or other morbidities. The Thai government must sort out this issue before it is too late as the whole nation’s health is at risk due to excessive dust levels higher than standard guidelines. Collection of pollution data can help us to come up with solutions and prevent it from turning into a hazardous situation. Unfortunately, pollution data are missing and need to be dealt with before analysis to obtain accurate results. Materials and methods: A new method of imputation for estimating population mean based on a transformed variable has been suggested under simple random sampling without replacement and the uniform nonresponse mechanism. The bias and mean square error of the proposed estimator are investigated up to the first order of approximation. The performance of the proposed estimator is studied via applications to air pollution data in Chiang Mai, Thailand. Results: The proposed estimator shows the best performance, giving the least bias and mean square error for all levels of sampling fractions. For the results from application the estimated value of sulfur dioxide from Particulate Matter 2.5 (PM2.5), the Percentage Relative Efficiency (PRE) is higher than all existing estimators by at least 16%. For the estimated PM2.5 from PM10 the PRE is higher than all existing estimators by at least 1600%, an extremely significant difference exhibiting similarity to real values. Conclusion: The proposed imputation technique based on the transformed auxiliary variable can be helpful for imputing missing values and improving the efficiency of the estimators.

Abstract 3000: A data-driven and AI-empowered systems biology model of MHC class I antigen presentation pathway

Abstract Immunotherapy is a new frontline therapeutic strategy for multiple cancer types. However, non-response mechanisms of immune checkpoint inhibitors generally exist in patients of solid cancer such breast and colorectal cancer. Multiple factors that parallelly regulate T cell functions may cause the non-response mechanism, such as regulators of T-cell’s recognition of cancer cells and cytotoxicity activities. Our recent study revealed that the presentation and salvage of MHC class I antigen presentation on cancer cells is a key deterministic factor of T-cell recognition in multiple cancer types. In this study, we established a systems biology model of the antigen processing and presentation pathways to identify cancer cell-expressing genes that are associated with immune response of breast cancer and effector T cell-mediated cytotoxicity. To the best of our knowledge, there is no established systems biology model of MHC I class antigen presentation pathway. We first manually collected and curated the biological processes that are involved in the MHC class I antigen presentation pathway via an extensive literature review. MHC I class antigen presentation involves a chain of biological processes, namely ubiquitination, proteosome, transporter associated with antigen processing, chaperone-based formation of MHC I class one complex, ER to Golgi, and Golgi to cell membrane exocytosis. We also include branches such as de-ubiquitination, MHC class I complex checking, and endocytosis-based salvage. We further estimated sample-wise activity level of the MHC class I antigen presentation pathway using transcriptomics data. By treating antigen as a “mass carrying” signal, we developed a systems biology model over the established pathway. The whole biological process was treated as a network model, in which the flux represents antigen presentation activity level. Based on this computational method, we further build a machine-learning pipeline that utilizes omics datasets (cancer genomics, scRNA-seq, spatial transcriptomics) collected from tumor tissue samples to correlate the MHC class I antigen presentation activities with immune response in cancer tumor microenvironment. We identified genes expressed by cancer cells that have high impacts on antigen presentation and are associated with T cell recognition and cytotoxicity. The identified genes are involved in (a) the antigen presentation machinery, (b) the immunoproteasome in antigen processing, (c) the dynamics of antigen presentation on the cell surface, and (d) cytokine and chemokine secretion. Citation Format: Jia Wang, Xiao Wang, Pengtao Dang, Haiqi Zhu, Xinyu Zhou, Kaman So, Sha Cao, Chi Zhang. A data-driven and AI-empowered systems biology model of MHC class I antigen presentation pathway [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 3000.

On representativeness, informative sampling, nonignorable nonresponse, semiparametric prediction and calibration

Informative sampling refers to a sampling design for which the sample selection probabilities depend on the values of the model outcome variable. In such cases the model holding for the sample data is different from the model holding for the population data. Similarly, nonignorable nonresponse refers to a nonresponse mechanism in which the response probability depends on the value of a missing outcome variable. For such a nonresponse mechanism the model holding for the response data is different from the model holding for the population data. In this paper, we study, within a modelling framework, the semi-parametric prediction of a finite population total by specifying the probability distribution of the response units under informative sampling and nonignorable nonresponse. This is the most general situation in surveys and other combinations of sampling informativeness and response mechanisms can be considered as special cases. Furthermore, based on the relationship between response distribution and population distribution, we introduce a new measure of the representativeness of a response set and a new test of nonignorable nonresponse and informative sampling, jointly. Finally, a calibration estimator is obtained when the sampling design is informative and the nonresponse mechanism is nonignorable.

Optogenetic stimulation of transmission from prelimbic cortex to nucleus accumbens core overcomes resistance to venlafaxine in an animal model of treatment-resistant depression

BackgroundOur earlier study demonstrated that repeated optogenetic stimulation of afferents from ventral hippocampus (vHIP) to the prelimbic region of medial prefrontal cortex (mPFC) overcame resistance to antidepressant treatment in Wistar-Kyoto (WKY) rats. These results suggested that antidepressant resistance may result from an insufficiency of transmission from vHIP to mPFC. Here we examined whether similar effects can be elicited from major output of mPFC; the pathway from to nucleus accumbens core (NAc). MethodWKY rats were subjected to Chronic Mild Stress and were used in two sets of experiments: 1) they were treated acutely with optogenetic stimulation of afferents to NAc core originating from the mPFC, and 2) they were treated with chronic (5 weeks) venlafaxine (10 mg/kg) and/or repeated (once weekly) optogenetic stimulation of afferents to NAc originating from either mPFC or vHIP. ResultsChronic mild stress procedure decreased sucrose intake, open arm entries on elevated plus maze, and novel object recognition test. Acute optogenetic stimulation of the mPFC-NAc and vHIP-NAc pathways had no effect in sucrose or plus maze tests, but increased object recognition. Neither venlafaxine nor mPFC-NAc optogenetic stimulation alone was effective in reversing the effects of CMS, but the combination of chronic antidepressant and repeated optogenetic stimulation improved behaviour on all three measures. ConclusionsThe synergism between venlafaxine and mPFC-NAc optogenetic stimulation supports the hypothesis that the mechanisms of non-responsiveness of WKY rats involves a failure of antidepressant treatment to restore transmission in the mPFC-NAc pathway. Together with earlier results, this implicates insufficiency in a vHIP-mPFC-NAc circuit in non-responsiveness to antidepressant drugs.

Neoadjuvant afatinib with paclitaxel for triple-negative breast cancer and the molecular characteristics in responders and non-responders

The prognosis of triple-negative breast cancer (TNBC) is worse and a major proportion of TNBC expresses epidermal growth factor receptor (EGFR). Afatinib can inhibit EGFR signal pathway; however, its treatment effect for TNBC is unknown. Thus, we aimed to assess the efficacy and biomarkers of afatinib in combination with paclitaxel in a neoadjuvant setting. Patients with stage II to III TNBC were enrolled. They received 40mg of afatinib daily for 14 days, followed by daily afatinib and weekly paclitaxel (80mg/m2) every 21 days for four to six cycles. To explore the mechanisms of responsiveness and non-responsiveness, 409 cancer-associated genes were sequenced. Twenty-one patients were enrolled and one patient achieved a complete clinical response; however, a 2mm residual tumor was noted in the surgical specimen. Overall, 33.0% patients were responders. Fifteen patients received molecular testing. No activated mutation of EGFR or Her2 were found. Activated PI3K or JAK2 pathway were trended to associate with non-responder (p=0.057). Mutation of homologous recombination (HR) genes were correlated with non-responsiveness (p=0.005). Seven patients did not have altered PI3K, JAK2 or HR pathway; six (85.7%) of them were responder. Patients with the amplified DAXX gene was associated with a favorable trend of response (p=0.109). Adding afatinib to neoadjuvant paclitaxel generated a modest effect in TNBC. Exploratory molecular analysis suggested that activated PI3K, JAK2 pathways and mutation of HR genes were associated with therapeutic non-responsiveness, and amplification of DAXX genes was associated with responsiveness to afatinib in combination with paclitaxel.

Analysis of Determinants on Item Nonresponse:A Study on Household/individual Income and Sexual Intercourse Nonresponse

This study examines the factors influencing item nonresponse, especially when it comes to income and sexual behavior, in the Japanese context. Many item nonresponses on questions on income and sexual behavior are a serious problem for analysis since they might cause biased estimations. However, many previous studies on item nonresponse when it comes to income and sexual issues examine surveys carried out in Western societies. Moreover, many previous studies have not distinguished between “Refused” and “Don't Know.” In addition, many previous studies on income nonresponse do not focus on the effect of religion, although questions on religion have been argued to affect survey cooperation. Thus, we analyzed the data from the Japanese General Social Survey 2000/2001 to distinguish the reason for nonresponse properly, reveal factors on income and sexual behavior nonresponses, and examined the effect of religion. We conducted multinomial logistic regressions of nonresponse in household income, individual income, and sexual behavior. Our findings are as follows: (1) Factors that affected nonresponses of income and sexual behavior are different between Japanese and Western societies. (2) The effect of age and educational background are different between “Refused” and “Don't Know.” (3) Respondents whose family has a religion tended to answer with household income. These findings indicate that we should consider the nonresponse mechanism due to the Japanese-specific context, distinguish the reason for nonresponse because of different effects of factors, and focus on the effect of religion on income nonresponse.

Single Imputation Methods and Confidence Intervals for the Gini Index

The problem of missing data is a common feature in any study, and a single imputation method is often applied to deal with this problem. The first contribution of this paper is to analyse the empirical performance of some traditional single imputation methods when they are applied to the estimation of the Gini index, a popular measure of inequality used in many studies. Various methods for constructing confidence intervals for the Gini index are also empirically evaluated. We consider several empirical measures to analyse the performance of estimators and confidence intervals, allowing us to quantify the magnitude of the non-response bias problem. We find extremely large biases under certain non-response mechanisms, and this problem gets noticeably worse as the proportion of missing data increases. For a large correlation coefficient between the target and auxiliary variables, the regression imputation method may notably mitigate this bias problem, yielding appropriate mean square errors. We also find that confidence intervals have poor coverage rates when the probability of data being missing is not uniform, and that the regression imputation method substantially improves the handling of this problem as the correlation coefficient increases.

Item-nonresponse in a household sample survey in Colombia

Item nonresponse occurs when sample units do not provide information on a particular variable, problem that may affect the representativeness of the sample and the reliability of the estimates. Efforts to reduce the item nonresponse rate do not necessarily improve the quality of the information. Besides the nonresponse rate, representativeness indicators can be used to measure the quality of the collected data. This paper analyzes the wage nonresponse mechanism of a household survey in Colombia and evaluates the quality of the wage information in two different periods of time (2008:4 and 2017:4). The results show a low but increasing nonresponse rate whose behavior does not seem to be associated with the set of observables considered. This result is of value since the selection of the adequate imputation method relies on the assumptions on the missing data mechanism.

Abstract A26: Humanized mouse model: A model to understand mechanisms of immune non-responsiveness to immune checkpoint inhibitors in melanoma

Abstract Immune checkpoint inhibitor therapy (anti-CTLA4 or anti-PD1 antibodies) is rapidly emerging as a front-line treatment option for many solid tumors. However, only a third of melanoma patients respond to immune checkpoint blockade. Currently available mouse xenograft and transgenic models have many shortcomings and are unable to address the basis of therapy resistance and immune nonresponsiveness that are observed in patients. Thus, there is an urgent need to establish an in vivo model with a human immune microenvironment that can address issues of therapy resistance. Our laboratory has developed a novel humanized mouse melanoma model. Immunodeficient NSG mice were reconstituted with human CD34+ cells and after 8-12 weeks, mice are fully reconstituted with human innate (monocyte/myeloid lineage cells, dendritic cells and NK cells) and adaptive (T and B cells) immune cells. Humanized mice were then challenged with HLA-matched melanoma cells, and the functional ability of human immune cells to restrict tumor growth was monitored. Delayed tumor growth was observed in humanized mice, indicating in vivo sensitization of human immune cells to melanoma. This was confirmed by in vitro demonstration of human lymphocytes from tumor-bearing mice showing enhanced cytokine expression after stimulation with melanoma antigen peptides. In therapy studies, tumor-bearing humanized mice treated with anti-PD-1 showed restricted tumor growth. Anti-PD-1 therapy resulted in enhanced infiltration of T cells that correlated with tumor response. MassCyTOF studies were performed using a panel of immune markers to understand the mechanism of therapy nonresponsiveness in some tumors. Downmodulation of HLA-class I molecules and increased presence of FOXP3+ cells in the tumor region were seen. Our results suggest that humanized mouse melanoma model can be explored further to understand the causes of therapy resistance and immune nonresponsiveness. Citation Format: Rajasekharan Somasundaram, Anastasia Samarkina, Thomas Connelly, Robin Choi, Hedy Choi, Kar Muthumani, Xiaowei Xu, Klaus Kaestner, Meenhard Herlyn. Humanized mouse model: A model to understand mechanisms of immune non-responsiveness to immune checkpoint inhibitors in melanoma [abstract]. In: Proceedings of the AACR Special Conference on Melanoma: From Biology to Target; 2019 Jan 15-18; Houston, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(19 Suppl):Abstract nr A26.

Sensitivity of five information criteria to discriminate covariance structures with missing data in repeated measures designs.

This study analyzes the effectiveness of different information criteria for the selection of covariance structures, extending it to different missing data mechanisms, the maintenance and adjustment of the mean structures, and matrices. The Monte Carlo method was used with 1,000 simulations, SAS 9.4 statistical software and a partially repeated measures design (p=2; q=5). The following variables were manipulated: a) the complexity of the model; b) sample size; c) matching of covariance matrices and sample size; d) dispersion matrices; e) the type of distribution of the variable; f) the non-response mechanism. The results show that all information criteria worked well in Scenario 1 for normal and non-normal distributions with heterogeneity of variance. However, in Scenarios 2 and 3, all were accurate with the ARH matrix, whereas AIC, AICCR and HQICR worked better with TOEP and UN. When the distribution was not normal, AIC and AICCR were only accurate in Scenario 3, more heterogeneous and unstructured matrices, with complete cases, MAR and MCAR. In order to correctly select the matrix it is advisible to analyze the heterogeneity, sample size and distribution of the data.

Parametric prediction of finite population total under Informative sampling and nonignorable nonresponse

Abstract In this paper, we combine two methodologies used in the model-based survey sampling, namely the prediction of the finite population total, named T, under informative sampling and full response, see Sverchkov and Pfeffermann (2004), and the prediction of T with a noninformative sampling design and the nonignorable nonresponse mechanism, see Eideh (2012). The former approach involves the dependence of the first order inclusion probabilities on the study variable, while the latter involves the dependence of the probability of nonresponse on unobserved or missing observations. The main aim of the paper is to consider how to account for the joint effects of informative sampling designs and not-missing-at-random response mechanism in statistical models for complex survey data. For this purpose, theoretically, we use the response distribution and relationships between the moments of the superpopoulation, the sample, sample-complement, response, and nonresponse distributions for the prediction of finite population totals, see Eideh (2016). The derived parametric predictors of T use the observation for the response set of the study variable or variable of interest, values of auxiliary variables and their population totals, sampling weights, and propensity scores. An interesting outcome of the T study is that most predictors known from model-based survey sampling can be derived as a special case from this general theory, see Chambers and Clark (2012).

Using Response Times for Joint Modeling of Response and Omission Behavior

For adequate modeling of missing responses, a thorough understanding of the nonresponse mechanisms is vital. As a large number of major testing programs are in the process or already have been moving to computer-based assessment, a rich body of additional data on examinee behavior becomes easily accessible. These additional data may contain valuable information on the processes associated with nonresponse. Bringing together research on item omissions with approaches for modeling response time data, we propose a framework for simultaneously modeling response behavior and omission behavior utilizing timing information for both. As such, the proposed model allows (a) to gain a deeper understanding of response and nonresponse behavior in general and, in particular, of the processes underlying item omissions in LSAs, (b) to model the processes determining the time examinees require to generate a response or to omit an item, and (c) to account for nonignorable item omissions. Parameter recovery of the proposed model is studied within a simulation study. An illustration of the model by means of an application to real data is provided.

Estimation tools for reducing the impact of sampling and nonresponse errors in dual-frame RDD telephone surveys.

We discuss alternative estimators of the population total given a dual-frame random-digit-dial (RDD) telephone survey in which samples are selected from landline and cell phone sampling frames. The estimators are subject to sampling and nonsampling errors. To reduce sampling variability when an optimum balance of landline and cell phone samples is not feasible, we develop an application of shrinkage estimation. We demonstrate the implications for survey weighting of a differential nonresponse mechanism by telephone status. We illustrate these ideas using data from the National Immunization Survey-Child, a large dual-frame RDD telephone survey sponsored by the Centers for Disease Control and Prevention and conducted to measure the vaccination status of American children aged 19 to 35months.

When Nonresponse Mechanisms Change: Effects on Trends and Group Comparisons in International Large-Scale Assessments.

Mechanisms causing item nonresponses in large-scale assessments are often said to be nonignorable. Parameter estimates can be biased if nonignorable missing data mechanisms are not adequately modeled. In trend analyses, it is plausible for the missing data mechanism and the percentage of missing values to change over time. In this article, we investigated (a) the extent to which the missing data mechanism and the percentage of missing values changed over time in real large-scale assessment data, (b) how different approaches for dealing with missing data performed under such conditions, and (c) the practical implications for trend estimates. These issues are highly relevant because the conclusions hold for all kinds of group mean differences in large-scale assessments. In a reanalysis of PISA (Programme for International Student Assessment) data from 35 OECD countries, we found that missing data mechanisms and numbers of missing values varied considerably across time points, countries, and domains. In a simulation study, we generated data in which we allowed the missing data mechanism and the amount of missing data to change over time. We showed that the trend estimates were biased if differences in the missing-data mechanisms were not taken into account, in our case, when omissions were scored as wrong, when omissions were ignored, or when model-based approaches assuming a constant missing data mechanism over time were used. The results suggest that the most accurate estimates can be obtained from the application of multiple group models for nonignorable missing values when the amounts of missing data and the missing data mechanisms changed over time. In an empirical example, we furthermore showed that the large decline in PISA reading literacy in Ireland in 2009 was reduced when we estimated trends using missing data treatments that accounted for changes in missing data mechanisms.

DEVELOPMENT OF CHAIN-TYPE EXPONENTIAL ESTIMATORS FOR POPULATION VARIANCE IN TWO-PHASE SAMPLING DESIGN IN PRESENCE OF RANDOM NON-RESPONSE

Abstract In this paper, an investigation has been carried out to deal with a unified approach of estimation procedures of population variance in two-phase sampling design under missing at random non-response mechanism circumstances. Using two auxiliary variables, we have developed different chain-type exponential estimators of finite population variance for two different set-ups and studied their properties under the different assumption of random non-response considered by Tracy and Osahan (1994). The comparisons of the proposed estimators have been made with some contemporary estimators of population variance under the similar realistic conditions. Numerical illustrations are presented to support the theoretical results. Results are analysed and suitable recommendations are put forward to the survey statisticians.

Sensitivity Analysis of Missing Mechanisms by using Multiple Imputation

다양한 자료에서 무응답이 관측되며, 이에 따라 무응답 처리에 대한 관심이 높아지고 있다. 무응답을 추정할 때는 무응답의 발생이 어떤 무응답 메커니즘을 따르는지에 대한 판단이 선행되어야 한다. 그러나 관측된 자료를 통해서는 무응답 메커니즘을 파악할 수 없다는 연구 결과에 따라, 다양한 무응답 메커니즘을 반영하여 무응답을 추정하고 이를 비교분석하는 민감도 분석을 수행할 것이 권고되고 있다. 본 논문에서는 다중대체법을 이용한 민감도 분석을 소개하고, 이를 우리나라 19대 대선여론조사 자료에 적용하였다. 각 지역의 투표 성향이 다르기 때문에 지역별 독립을 가정하고, 지역별로 각각 민감도 분석을 실시하였다. 무응답 추정을 위한 모형으로는 다범주 로지스틱 모형을 이용하였으며, 민감도 모수의 변화에 따른 문재인 후보 지지율의 변화를 살펴보았다. 분석 결과 일부 지역에서는 개표 결과가 MAR(missing at random) 메커니즘 가정 하에서 계산된 신뢰구간의 밖에 존재함을 알 수 있었으며, 타 지역의 경우에도 MAR 메커니즘 가정 하에서의 결과보다 실제 개표결과에 더 가까운 추정치가 존재하였다. 본 연구의 결과가 다양한 선거 자료에 적용되어 민감도 분석 결과가 축적된다면, 향후 선거예측의 정밀도를 보다 높일수 있을 것으로 기대된다.Nonresponse is observed in various data, and the importance of handling nonresponses is increasing. When estimating nonresponse, it is necessary to determine which missing mechanism result in the occurrence of nonresponses. Due to the recent research result of non-identifiability of the missing mechanism, it is recommended to perform sensitivity analysis which compare estimates from various non-response mechanisms. In this paper, we introduce the sensitivity analysis using the multiple imputation method and applied it to the 19th Korean national poll survey data. In the sensitivity analysis which use logistic regression model with multiple categories under the regional independence, and the changes of supporting rates of the candidate Moon was investigated with respect to the changes of the sensitivity parameters. As a result, for a certain region, the supporting rate of the election result located outside of the confidence interval under the assumption of the MAR (missing at random) mechanism. In other regions, estimates under other mechanism are closer to the election result. If the results of this study are applied to various election data, it is expected that the accuracy of future election prediction will be improved.

Calibration Weighting for Nonresponse with Proxy Frame Variables (So that Unit Nonresponse Can Be Not Missing at Random)

Abstract When adjusting for unit nonresponse in a survey, it is common to assume that the response/nonresponse mechanism is a function of variables known either for the entire sample before unit response or at the aggregate level for the frame or population. Often, however, some of the variables governing the response/nonresponse mechanism can only be proxied by variables on the frame while they are measured (more) accurately on the survey itself. For example, an address-based sampling frame may contain area-level estimates for the median annual income and the fraction home ownership in a Census block group, while a household’s annual income category and ownership status are reported on the survey itself for the housing units responding to the survey. A relatively new calibration-weighting technique allows a statistician to calibrate the sample using proxy variables while assuming the response/ nonresponse mechanism is a function of the analogous survey variables. We will demonstrate how this can be done with data from the Residential Energy Consumption Survey National Pilot, a nationally representative web-and-mail survey of American households sponsored by the U.S. Energy Information Administration.