Methods for Imputing Missing Values and Synthesizing Confidential Values for Continuous and Magnitude Data

Abstract

Continuous variable is one of the major data types collected by the survey organizations. It can be incomplete such that the data collectors need to fill in the missingness. Or, it can contain sensitive information which needs protection from re-identification. One of the approaches to protect continuous microdata is to sum them up according to different cells of features. In this thesis, I represents novel methods of multiple imputation (MI) that can be applied to impute missing values and synthesize confidential values for continuous and magnitude data.

The first method is for limiting the disclosure risk of the continuous microdata whose marginal sums are fixed. The motivation for developing such a method comes from the magnitude tables of non-negative integer values in economic surveys. I present approaches based on a mixture of Poisson distributions to describe the multivariate distribution so that the marginals of the synthetic data are guaranteed to sum to the original totals. At the same time, I present methods for assessing disclosure risks in releasing such synthetic magnitude microdata. The illustration on a survey of manufacturing establishments shows that the disclosure risks are low while the information loss is acceptable.

The second method is for releasing synthetic continuous micro data by a nonstandard MI method. Traditionally, MI fits a model on the confidential values and then generates multiple synthetic datasets from this model. Its disclosure risk tends to be high, especially when the original data contain extreme values. I present a nonstandard MI approach conditioned on the protective intervals. Its basic idea is to estimate the model parameters from these intervals rather than the confidential values. The encouraging results of simple simulation studies suggest the potential of this new approach in limiting the posterior disclosure risk.

The third method is for imputing missing values in continuous and categorical variables. It is extended from a hierarchically coupled mixture model with local dependence. However, the new method separates the variables into non-focused (e.g., almost-fully-observed) and focused (e.g., missing-a-lot) ones. The sub-model structure of focused variables is more complex than that of non-focused ones. At the same time, their cluster indicators are linked together by tensor factorization and the focused continuous variables depend locally on non-focused values. The model properties suggest that moving the strongly associated non-focused variables to the side of focused ones can help to improve estimation accuracy, which is examined by several simulation studies. And this method is applied to data from the American Community Survey.

No-Holdback allocation rules for continuous-time assemble-to-order systems

This paper analyzes a class of common-component allocation rules, termed no-holdback (NHB) rules, in continuous-review assemble-to-order (ATO) systems with positive lead times. The inventory of each component is replenished following an independent base-stock policy. In contrast to the usually assumed first-come-first-served (FCFS) component allocation rule in the literature, an NHB rule allocates a component to a product demand only if it will yield immediate fulfillment of that demand. We identify metrics as well as cost and product structures under which NHB rules outperform all other component allocation rules. For systems with certain product structures, we obtain key performance expressions and compare them to those under FCFS. For general product structures, we present performance bounds and approximations. Finally, we discuss the applicability of these results to more general ATO systems. © 2010 INFORMS.

Carbon mitigation costs for the commercial building sector: Discrete-continuous choice analysis of multifuel energy demand

We estimate a carbon mitigation cost curve for the U.S. commercial sector based on econometric estimation of the responsiveness of fuel demand and equipment choices to energy price changes. The model econometrically estimates fuel demand conditional on fuel choice, which is characterized by a multinomial logit model. Separate estimation of end uses (e.g., heating, cooking) using the U.S. Commercial Buildings Energy Consumption Survey allows for exceptionally detailed estimation of price responsiveness disaggregated by end use and fuel type. We then construct aggregate long-run elasticities, by fuel type, through a series of simulations; own-price elasticities range from -0.9 for district heat services to -2.9 for fuel oil. The simulations form the basis of a marginal cost curve for carbon mitigation, which suggests that a price of $20 per ton of carbon would result in an 8% reduction in commercial carbon emissions, and a price of $100 per ton would result in a 28% reduction. © 2008 Elsevier B.V. All rights reserved.

Research Question, Study Design and Continuous Research Education and Training Exercises (CREATE) Program.

The clinical research project starts with identifying the optimal research question, one that is ethical, impactful, feasible, scientifically sound, novel, relevant, and interesting. The project continues with the design of the study to answer the research question. Such design should be consistent with ethical and methodological principles, and make optimal use of resources in order to have the best chances of identifying a meaningful answer to the research question. Physicians and other healthcare providers are optimally positioned to identify meaningful research questions the answer to which could make significant impact on healthcare delivery. The typical medical education curriculum, however, lacks solid training in clinical research. We propose CREATE (Continuous Research Education And Training Exercises) as a peer- and group-based, interactive, analytical, customized, and accrediting program with didactic, training, mentoring, administrative, and professional support to enhance clinical research knowledge and skills among healthcare professionals, promote the generation of original research projects, increase the chances of their successful completion and potential for meaningful impact. The key features of the program are successive intra- and inter-group discussions and confrontational thematic challenges among participating peers aimed at capitalizing on the groups' collective knowledge, experience and skills, and combined intellectual processing capabilities to optimize choice of research project elements and stakeholder decision-making.

Bayesian Gaussian Copula Factor Models for Mixed Data.

Gaussian factor models have proven widely useful for parsimoniously characterizing dependence in multivariate data. There is a rich literature on their extension to mixed categorical and continuous variables, using latent Gaussian variables or through generalized latent trait models acommodating measurements in the exponential family. However, when generalizing to non-Gaussian measured variables the latent variables typically influence both the dependence structure and the form of the marginal distributions, complicating interpretation and introducing artifacts. To address this problem we propose a novel class of Bayesian Gaussian copula factor models which decouple the latent factors from the marginal distributions. A semiparametric specification for the marginals based on the extended rank likelihood yields straightforward implementation and substantial computational gains. We provide new theoretical and empirical justifications for using this likelihood in Bayesian inference. We propose new default priors for the factor loadings and develop efficient parameter-expanded Gibbs sampling for posterior computation. The methods are evaluated through simulations and applied to a dataset in political science. The models in this paper are implemented in the R package bfa.

Protocol dependence of the jamming transition.

We propose a theoretical framework for predicting the protocol dependence of the jamming transition for frictionless spherical particles that interact via repulsive contact forces. We study isostatic jammed disk packings obtained via two protocols: isotropic compression and simple shear. We show that for frictionless systems, all jammed packings can be obtained via either protocol. However, the probability to obtain a particular jammed packing depends on the packing-generation protocol. We predict the average shear strain required to jam initially unjammed isotropically compressed packings from the density of jammed packings, shape of their basins of attraction, and path traversed in configuration space. We compare our predictions to simulations of shear strain-induced jamming and find quantitative agreement. We also show that the packing fraction range, over which shear strain-induced jamming occurs, tends to zero in the large system limit for frictionless packings with overdamped dynamics.

Differences in onset and abuse/dependence episodes between prescription opioids and heroin: results from the National Epidemiologic Survey on Alcohol and Related Conditions.

OBJECTIVES: To examine patterns of onset and abuse/dependence episodes of prescription opioid (PO) and heroin use disorders in a national sample of adults, and to explore differences by gender and substance abuse treatment status. METHODS: Analyses of data from the 2001-2002 National Epidemiologic Survey on Alcohol and Related Conditions (N = 43,093). RESULTS: Of all respondents, 5% (n = 1815) reported a history of nonmedical PO use (NMPOU) and 0.3% (n = 150) a history of heroin use. Abuse was more prevalent than dependence among NMPOUs (PO abuse, 29%; dependence, 7%) and heroin users (heroin abuse, 63%; dependence, 28%). Heroin users reported a short mean interval from first use to onset of abuse (1.5 years) or dependence (2.0 years), and a lengthy mean duration for the longest episode of abuse (66 months) or dependence (59 months); the corresponding mean estimates for PO abuse and dependence among NMPOUs were 2.6 and 2.9 years, respectively, and 31 and 49 months, respectively. The mean number of years from first use to remission from the most recent episode was 6.9 years for PO abuse and 8.1 years for dependence; the mean number of years from first heroin use to remission from the most recent episode was 8.5 years for heroin abuse and 9.7 years for dependence. Most individuals with PO or heroin use disorders were remitted from the most recent episode. Treated individuals, whether their problem was heroin or POs, tended to have a longer mean duration of an episode than untreated individuals. CONCLUSION: Periodic remissions from opioid or heroin abuse or dependence episodes occur commonly but take a long time. Timely and effective use of treatment services are needed to mitigate the many adverse consequences from opioid/heroin abuse and dependence.