Analysis

How do we tell whether weighting adjustments reduce nonresponse bias? If a variable is measured for everyone in the selected sample, then the design weights can be used to calculate an approximately unbiased estimate of the population mean or total for that variable. A second estimate of the population mean or total can be calculated using the survey respondents only, with weights that have been adjusted for nonresponse. If the two estimates disagree, then there is evidence that the weight adjustments may not have removed the nonresponse bias for that variable. In this paper we develop the theoretical properties of linearization and jackknife variance estimators for evaluating the bias of an estimated population mean or total by comparing estimates calculated from overlapping subsets of the same data with different sets of weights, when poststratification or inverse propensity weighting is used for the nonresponse adjustments to the weights. We provide sufficient conditions on the population, sample, and response mechanism for the variance estimators to be consistent, and demonstrate their small-sample properties through a simulation study.

Designs and estimators for the single frame surveys currently used by U.S. government agencies were developed in response to practical problems. Federal household surveys now face challenges of decreasing response rates and frame coverage, higher data collection costs, and increasing demand for small area statistics. Multiple frame surveys, in which independent samples are drawn from separate frames, can be used to help meet some of these challenges. Examples include combining a list frame with an area frame or using two frames to sample landline telephone households and cellular telephone households. We review point estimators and weight adjustments that can be used to analyze multiple frame surveys with standard survey software, and summarize construction of replicate weights for variance estimation. Because of their increased complexity, multiple frame surveys face some challenges not found in single frame surveys. We investigate misclassification bias in multiple frame surveys, and propose a method for correcting for this bias when misclassification probabilities are known. Finally, we discuss research that is needed on nonsampling errors with multiple frame surveys.

Gross flows are often used to study transitions in employment status or other categorical variables among individuals in a population. Dual frame longitudinal surveys, in which independent samples are selected from two frames to decrease survey costs or improve coverage, can present challenges for efficient and consistent estimation of gross flows because of complex designs and missing data in either or both samples. We propose estimators of gross flows in dual frame surveys and examine their asymptotic properties. We then estimate transitions in employment status using data from the Current Population Survey and the Survey of Income and Program Participation.

This paper examines a jackknife method proposed in Rao (2003) for estimating mean squared errors (MSEs) when generalized linear models or other non-linear models are used for the response of interest. It demonstrate the method's performance in a simulation study.

In the United States, the National Health and Nutrition Examination Survey (NHANES) is linked to the National Health Interview Survey (NHIS) at the primary sampling unit level (the same counties, but not necessarily the same persons, are in both surveys). The NHANES examines about 5,000 persons per year, while the NHIS samples about 100,000 persons per year. In this paper, we present and develop properties of models that allow NHIS and administrative data to be used as auxiliary information for estimating quantities of interest in the NHANES. The methodology, related to Fay-Herriot (1979) small-area models and to calibration estimators in Deville and Sarndal (1992), accounts for the survey designs in the error structure.

How do we tell whether weighting adjustments reduce nonresponse bias? If a variable is measured for everyone in the selected sample, then the design weights can be used to calculate an approximately unbiased estimate of the population mean or total for that variable. A second estimate of the population mean or total can be calculated using the survey respondents only, with weights that have been adjusted for nonresponse. If the two estimates disagree, then there is evidence that the weight adjustments may not have removed the nonresponse bias for that variable. In this paper we develop the theoretical properties of linearization and jackknife variance estimators for evaluating the bias of an estimated population mean or total by comparing estimates calculated from overlapping subsets of the same data with different sets of weights, when poststratification or inverse propensity weighting is used for the nonresponse adjustments to the weights. We provide sufficient conditions on the population, sample, and response mechanism for the variance estimators to be consistent, and demonstrate their small-sample properties through a simulation study.

Designs and estimators for the single frame surveys currently used by U.S. government agencies were developed in response to practical problems. Federal household surveys now face challenges of decreasing response rates and frame coverage, higher data collection costs, and increasing demand for small area statistics. Multiple frame surveys, in which independent samples are drawn from separate frames, can be used to help meet some of these challenges. Examples include combining a list frame with an area frame or using two frames to sample landline telephone households and cellular telephone households. We review point estimators and weight adjustments that can be used to analyze multiple frame surveys with standard survey software, and summarize construction of replicate weights for variance estimation. Because of their increased complexity, multiple frame surveys face some challenges not found in single frame surveys. We investigate misclassification bias in multiple frame surveys, and propose a method for correcting for this bias when misclassification probabilities are known. Finally, we discuss research that is needed on nonsampling errors with multiple frame surveys.

Gross flows are often used to study transitions in employment status or other categorical variables among individuals in a population. Dual frame longitudinal surveys, in which independent samples are selected from two frames to decrease survey costs or improve coverage, can present challenges for efficient and consistent estimation of gross flows because of complex designs and missing data in either or both samples. We propose estimators of gross flows in dual frame surveys and examine their asymptotic properties. We then estimate transitions in employment status using data from the Current Population Survey and the Survey of Income and Program Participation.

This paper examines a jackknife method proposed in Rao (2003) for estimating mean squared errors (MSEs) when generalized linear models or other non-linear models are used for the response of interest. It demonstrate the method's performance in a simulation study.

In the United States, the National Health and Nutrition Examination Survey (NHANES) is linked to the National Health Interview Survey (NHIS) at the primary sampling unit level (the same counties, but not necessarily the same persons, are in both surveys). The NHANES examines about 5,000 persons per year, while the NHIS samples about 100,000 persons per year. In this paper, we present and develop properties of models that allow NHIS and administrative data to be used as auxiliary information for estimating quantities of interest in the NHANES. The methodology, related to Fay-Herriot (1979) small-area models and to calibration estimators in Deville and Sarndal (1992), accounts for the survey designs in the error structure.