Search

In sample surveys that collect information on skewed variables, it is often desirable to assess the influence of sample units on the sampling error of survey-weighted estimators of finite population parameters. The conditional bias is an attractive measure of influence that accounts for the sampling design and the estimation method. It is defined as the design expectation of the sampling error conditional on a given unit being selected in the sample. The estimation of the conditional bias is relatively straightforward for simple sampling designs and estimators. However, for complex designs or complex estimators, it may be tedious to derive an explicit expression for the conditional bias. In those complex surveys, variance estimation is often achieved through replication methods such as the bootstrap. Bootstrap methods of variance estimation are typically implemented by producing a set of bootstrap weights that is provided to users along with the survey data. In this article, we show how to use these bootstrap weights to obtain an estimator of the conditional bias. Our bootstrap estimator is evaluated in a simulation study and illustrated using data from the Canadian Survey of Household Spending. [ABSTRACT FROM AUTHOR]

Differential privacy is a restriction on data processing algorithms that provides strong confidentiality guarantees for individual records in the data. However, research on proper statistical inference, that is, research on properly quantifying the uncertainty of the (noisy) sample estimate regarding the true value in the population, is currently still limited. This article proposes and evaluates several strategies to compute valid differentially private confidence intervals for the median. Instead of computing a differentially private point estimate and deriving its uncertainty, we directly estimate the interval bounds and discuss why this approach is superior if ensuring privacy is important. We also illustrate that addressing both sources of uncertainty—the error from sampling and the error from protecting the output—simultaneously should be preferred over simpler approaches that incorporate the uncertainty in a sequential fashion. We evaluate the performance of the different algorithms under various parameter settings in extensive simulation studies and demonstrate how the findings could be applied in practical settings using data from the 1940 Decennial Census. [ABSTRACT FROM AUTHOR]

The National Marine Fisheries Service is the agency charged with estimating the number of fish removed from US oceans by recreational anglers. Two surveys produce data for these estimates; one measures number of angler trips and the other fish caught per trip by species, geography, and time period. Both surveys collect data from people and have the usual sources of nonsampling error afflicting demographic surveys. Due to lack of accessibility to fishing sites, they also have additional sources such as undercoverage. The rare incidence of saltwater fishing among the general population makes the current method costly and time-consuming. Consequently, new ways of obtaining information to supplement or replace the current method are of interest. One is the electronic logbook (ELB); this approach allows anglers to self-report their catch using cellphones or other communication devices. Estimation of catch from these data are possible with estimators using capture-recapture methods, but new sources of nonsampling error are rising. In this article, we examine three sources of nonsampling error in estimators of catch and approximate their biasing effect. We illustrate the method by comparing the effects using data from an ELB study in the Gulf of Mexico. [ABSTRACT FROM AUTHOR]

In the analysis of survey data, it is of interest to estimate and quantify uncertainty about means or totals for each of several nonoverlapping subpopulations or areas. When the sample size for a given area is small, standard confidence intervals based on data only from that area can be unacceptably wide. In order to reduce interval width, practitioners often utilize multilevel models in order to borrow information across areas, resulting in intervals centered around shrinkage estimators. However, such intervals only have the nominal coverage rate on average across areas under the assumed model for across-area heterogeneity. The coverage rate for a given area depends on the actual value of the area mean and can be nearly zero for areas with means that are far from the across-group average. As such, the use of uncertainty intervals centered around shrinkage estimators are inappropriate when area-specific coverage rates are desired. In this article, we propose an alternative confidence interval procedure for area means and totals under normally distributed sampling errors. This procedure not only has constant 1 − α frequentist coverage for all values of the target quantity but also uses auxiliary information to borrow information across areas. Because of this, the corresponding intervals have shorter expected lengths than standard confidence intervals centered on the unbiased direct estimator. Importantly, the coverage of the procedure does not depend on the assumed model for across-area heterogeneity. Rather, improvements to the model for across-area heterogeneity result in reduced expected interval width. [ABSTRACT FROM AUTHOR]

Although random routes are an important face-to-face sampling method, there is very little research about the procedure. The idea that random route sampling leads to a uniform selection of households has never been tested. This study challenges this notion in a simulation and uses registration office data to verify the impact of selection errors on survey results. All tested routes strongly violate the equal probability assumption and lead to biased expected values in multiple variables. The strongest errors were found in variables that were related to the spatial location of households. Furthermore, errors were related not only to random route instructions but also to the selection of starting points for these routes. This study is of relevance for scientists using random route data for their research as well as for surveys applying random route procedures. Data analysts can use the results to gain an impression about the degree of bias, whereas researchers that use random route samples should view these findings as inducement to improve the random route methodology and develop alternatives. [ABSTRACT FROM AUTHOR]