Elizabeth A. Stasny

Hierarchical Models for the Probabilities of a Survey Classification and Nonresponse: An Example from the National Crime Survey

Abstract A goal in many survey sampling problems is to estimate the probability that elements of the population within various small areas or domains have some characteristic or fall in some particular survey classification. The estimation problem is typically complicated by nonrandom nonresponse in that the probability that a unit responds to the survey may be related to the characteristic of interest. This article presents a random parameter or hierarchical model approach to modeling the small-domain probabilities of the characteristic of interest and the probabilities of nonresponse. The general model allows nonresponse probabilities to depend on a units survey classification. A special case of the model treats nonresponse as occurring at random. Empirical Bayes methods are used to obtain parameter estimates under the hierarchical models. The method is illustrated using data from the National Crime Survey.

Estimating Gross Flows Using Panel Data with Nonresponse: An Example from the Canadian Labour Force Survey

Abstract This article considers the problem of using categorical data from a panel survey in which there is nonrandom nonresponse to estimate gross flows. The methods are illustrated for the case of estimating gross flows in labor force participation using data from the Canadian Labour Force Survey. Three models are proposed that allow nonresponse to be related to employment classification, time, or both employment classification and time. Maximum likelihood estimation is used to fit the models to a single panel of Labour Force Survey data.

An application of ranked set sampling for mean and median estimation using USDA crop production data

Ranked set sampling (RSS) is a sampling approach that leads to improved statistical inference in situations where the units to be sampled can be ranked (either through some subjective judgment or via the use of an auxiliary variable) relative to each other prior to formal measurement. It has the most promise for leading to improved methodology in situations where ranking of the items to be sampled can be carried out relatively easily and cheaply compared to the effort and expense required for actual quantification of the characteristic of interest. Although the theoretical benefits of RSS in estimation and statistical inference have been extensively demonstrated in the literature, the methodology has not yet been widely adopted by practitioners. The aim of this study is to use a crop production dataset from the United States Department of Agriculture to demonstrate the practical benefits of RSS relative to the more commonly used simple random sampling in estimation of the mean and median of a population. The results of our study provide clear evidence that the use of RSS can lead to substantial gains in precision of estimation for both of these situations.

Relative effects of high saturated fatty acid levels in meat, dairy products, and tropical oils on serum lipoproteins and low-density lipoprotein degradation by mononuclear cells in healthy males

To determine the effects of three saturated fatty acid combinations on lipoprotein metabolism, we fed 18 21- to 32-year-old men three diets in a crossover design for 28-day periods separated by washout periods of 4 to 6 weeks. The men self-selected a prescribed diet at home emphasizing saturated fat as the visible fat for 1 week. Then, they ate experimental diets providing 40%, 15%, 17%, and 7% of food energy, respectively, as total, saturated, monounsaturated, and polyunsaturated fatty acids, levels representing amounts available in the US diet. Different test fatty acid combinations, given at 4 to 6 energy% (en%) each, were incorporated into food products: 12:0 + 14:0, 14:0 + 16:0, and 16:0 + 18:0. Test fatty acids were equalized by giving free myristic acid (14:0) with palm kernel oil or butter and sheanut butter (high in 18:0) with lard. The diet highest in 12:0 + 14:0 also provided 4.2 en% 16:0, the most common saturated fatty acid in the US diet. Mean apparent absorption of all fatty acids was at least 90%. The three diets produced similar concentrations of serum total and low-density lipoprotein (LDL) cholesterol and apolipoprotein (apo) B-100 regardless of the apo E phenotype of the subjects. Compared with baseline, the experimental diets affected serum high-density lipoprotein (HDL) concentrations (P < .06), with the highest values occurring on diet 12:0 + 14:0. When the change from baseline in receptor-mediated degradation of 125I-LDL in freshly isolated mononuclear cells (MNC) was stratified by apo E phenotype, diet 16:0 + 18:0 produced a 30% increase, compared with a 9% decrease on diet 12:0 + 14:0 and a 6% increase on diet 14:0 + 16:0 in subjects with the apo E3/3 phenotype. These results suggested that different saturated fatty acid combinations, consumed at levels typical of availability in the United States and with diets providing ample unsaturated fat, had similar cholesterolemic properties in healthy males despite some subtly different effects on lipoprotein metabolism.

Cautionary note on unbalanced ranked-set sampling

Balanced ranked-set sampling (RSS) offers improved statistical inference in situations where the units to be sampled can be ranked relative to each other prior to formal measurement. Recent work has shown that provided the ranking process is perfect, unbalanced RSS can do even better. In this article, we examine the performance of one unbalanced RSS technique when the ranking process is not perfect. Using an Ohio corn production data set, we show that median-based unbalanced RSS outperforms balanced RSS in estimating a population median if the rankings are nearly perfect. We also show, however, that median-based unbalanced RSS may perform extremely poorly when the ranking process is less than perfect. This effect is particularly pronounced when the variable of interest has a skewed distribution. We thus offer a note of caution for users of unbalanced RSS.

Teaching Survey Sampling

Abstract In recent years the focus of research in survey sampling has changed to include a number of nontraditional topics such as nonsampling errors. In addition, the availability of data from large-scale sample surveys, along with computers and software to analyze the data, have changed the tools needed by survey sampling statisticians. It has also resulted in a diverse group of secondary data users who wish to learn how to analyze data from a complex survey. Thus it is time to reassess what we should be teaching students about survey sampling. This article brings together a panel of experts on survey sampling and teaching to discuss their views on what should be taught in survey sampling classes and how it should be taught.

Modeling Nonignorable Nonresponse in Categorical Panel Data with an Example in Estimating Gross Labor-Force Flows

Many large-scale sample surveys use panel designs under which sampled individuals are interviewed several times before being dropped from the sample. The longitudinal data bases available from such surveys could be used to provide estimates of gross change over time. One problem in using these data to estimate gross change is how to handle the period-to-period nonresponse. This nonresponse is typically nonrandom and, furthermore, may be nonignorable in that it cannot be accounted for by other observed quantities in the data. Under the models proposed in this article, which are appropriate for the analysis of categorical data, the probability of nonresponse may be taken to be a function of the missing variable of interest. The proposed models are fit using maximum likelihood estimation. As an example, the method is applied to the problem of estimating gross flows in labor-force participation using data from the Current Population Survey and the Canadian Labour Force Survey.

Unbalanced Ranked Set Sampling for Estimating A Population Proportion Under Imperfect Rankings

The application of unbalanced ranked set sampling (RSS) to estimation of a population proportion has been studied for the perfect ranking situation. When the rankings are not perfect, the probabilities of success ranks for the judgment order statistics incorporate information on ranking errors as well as ranks. The objective of this article is to investigate the ranking errors effect of imperfection in rankings on unbalanced RSS for binary variables and provide methods to obtain estimates for the probabilities of success for the judgment order statistics using training samples so that Neyman allocation can be implemented. We also use a substantial data set, the NHANES III data, to demonstrate the feasibility and benefits of Neyman allocation in RSS for binary variables in the case of imperfect rankings.

How to Get a Job in Academics

Doctoral students in statistics who seek jobs in academics may know very little about the process of obtaining such a job and what they should do during their years in graduate school to improve their chances of being offered a job in academics. Although students may see job candidates present seminars in their departments, they are unlikely to be aware of what the candidate is doing during the rest of the visit to the department. They are also unlikely to gain from the experiences of other students who have gone through the job search process in previous years because those students graduate and leave and, hence, are unavailable to give advice. This article presents information and advice for students who plan to seek a job in academics. Although some of the article pertains to the interview process itself, it also gives advice for preparing for a job in academics throughout the students years in graduate school. Much of this advice is applicable to all Ph.D. students, whether or not they go into academics. Although the advice is meant for graduate students, this article may also be a useful resource for advisors, mentors, department heads, and others who help guide students through their graduate studies.

Optimal ranked set sampling estimation based on medians from multiple set sizes

Ranked set sampling (RSS) is a sample selection technique that makes use of expert knowledge to rank sample units before measuring them. Even though rankings are not always perfect, RSS is useful in situations where obtaining measurements is costly, difficult, or destructive. Research in this area has tended to focus on the case where all set sizes are equal. This article represents a departure from that setting because we encounter different set sizes within a single sample. More specifically, we propose an alternative estimator for the median of a symmetric distribution using medians of ranked set samples of various set sizes from such a distribution. This estimator is seen to be robust over a wide class of symmetric distributions.