Monday, Aug 5: 10:30 AM - 12:20 PM
5060
Contributed Papers
Oregon Convention Center
Room: CC-C125
Main Sponsor
Survey Research Methods Section
Presentations
Respondent-driven sampling (RDS) and RDS-like methods, such as Link-Tracing Sampling and Network Sampling, have been widely used in studies of hidden or hard-to-reach populations over the past decade. However, little or no literature has addressed the issue of the violations of RDS or network sampling assumptions and data deficiencies, a challenge that are frequently encountered during RDS implementation due to the uncertainties of fieldwork. To this end, we present an empirical application of a novel estimate for the RDS or RDS-like sampling methods called new estimates for network sampling (NE4NS). It is resampling based, free of RDS and model-based assumptions, as opposed to the conventional RDS estimate in which the Volz-Heckathorn (VH) weighting scheme relies on the self-reported network sizes. The new and the conventional RDS estimations were applied to a sex trafficking prevalence study in Senegal, one of our ongoing projects on which the RDS method was used and the problem of insufficient sample size was faced. As a result, the new RDS estimate with the NE4NS strategy showed to be highly efficient and effective. Future application of the new RDS estimate is encouraged.
Keywords
Respondent-Driven Sampling (RDS)
Network Sampling
New Estimates for Network Sampling (NE4NS)
Volz-Heckathorn (VH) Weighting Scheme
Hidden or Hard-to-Reach Population
Resampling Approach
The apportionment of House seats to states after each decennial census can be viewed as a probability proportional to size (pps) sample allocation to the states. The sample size is 435, and the measure of size is the state's apportionment population (resident population plus U.S. federal employees and dependents living overseas, allocated to their home state), subject to the constraint that all states must receive at least one seat. An investigation of a group of sample allocations to areas that were supposed to be pps led to the discovery that the current method (Huntington-Hill) used for apportionment assigned excessive sample to areas with larger populations in the early sample assignment stages. An alternative algorithm was constructed that assigned sample to areas by minimizing the sum of the absolute values of the difference between the population proportion and the sample proportion across the areas at each stage. Subsequently the alternative algorithm was used to do a hypothetical apportionment following each census from 1790 and 2020. The results are similar or identical to Webster's method (used after the 1840, 1910, 1930 censuses).
Keywords
Sample allocation
Recently the National Health Interview Survey tested the cycling of non-self-representing (NSR) Primary Sample Units in six states. A pilot study was conducted during the 2010-based design period to test operational feasibility and to evaluate the reliability of multi-year state estimates where the sample of NSR first stage units is not the same in all years. The pilot study was considered a success, and the National Center for Health Statistics requested that the Census Bureau include the cycling of NSR first stage units as a design feature for all states with NSR units in the upcoming sample redesign based on the 2020 Decennial Census. The primary goal of the cycling is to ensure sufficient degrees of freedom to allow reliable estimates for each state when pooling three years of data. In planning this sample redesign, we initially included a maximal set of counties in the overall cycling sample for each state. However, we found that this approach was neither necessary nor desirable to meet the requirements set forth by the survey sponsor. In this paper we explain how we arrived at our final first stage sample design and describe how the cycling of the NSR units is carried out.
Keywords
Survey Sample Design, Multi-stage Sampling, Primary Sample Units, Cycling, National Health Interview Survey
Abstract
The Current Population Survey (CPS) conducts data collection monthly using an eight-panel rotating sample. Field operations and headquarters both require a predictable sample size for planning purposes. However, various factors impact the sample size of each panel; among these factors are growth in the frame which gets added to the panels differently across time, attrition in the frame which causes a decrease in the eligible sample size across time, and post-census frame improvements which reduce that decrease in our eligible sample. In our research, we first discuss the components of the CPS sample size. Next, we attempt to model these components so that we can extrapolate the sample size months ahead. Lastly, we compare our extrapolations to observed sample sizes to evaluate the quality of our predictions.
Key words: Sample Frames, Modeling
Keywords
Sample Frame
Modeling
Capture-Recapture methods are often used to estimate the size of populations. The models underlying these methods often require characteristics to be reported similarly across two independent measurements. For the 2020 Census and Post-Enumeration Survey there was meaningful disagreement in the measurement of occupancy, race, and Hispanic Origin. For example, of the 12,400 vacant housing units in a sample of census housing units that matched to the Post-Enumeration Survey, 50% were classified as occupied by the Post-Enumeration Survey. In this study, we report the degree of inconsistency between the two independent measurements and analyze how measurement error impacts dual-system estimates of the sub-population sizes.
Keywords
Measurement Error
Census
Coverage
Dual-System Estimation
Capture-Recapture
Every other November, Americans vote on an array of candidates for federal, state, and local offices. Although many offices are on the ballot each election, voter behavior increasingly appears low-dimensional due to the dominant two-party system and rising partisan polarization. Nevertheless, voter policy preferences remain more heterogeneous than may be appreciated based solely on candidate performance in partisan elections. As in several states, constitutional amendments in Florida require direct approval by the electorate via statewide referendum. From 2012-2022, 40 such proposed amendments have been voted on, spanning policy areas such as taxation, healthcare, and civil liberties. The corresponding results provide a rich resource to interrogate voter preferences directly across a spectrum of issues. Combining official precinct-level results and available socioeconomic and demographic data, this study characterizes the support for ballot initiatives and candidates in Florida using techniques for regression, dimension reduction, and deconvolution. To conclude, the methodological challenges inherent to this data type and implications of the findings are discussed.
Keywords
Election data
public opinion
decomposition
deconvolution
public policy
regression