Suchergebnisse

In this note, we study how parameter vector estimation for a trigonometric regression model and the expected squared residual error computed from an estimated model are affected by Berkson‐type measurement error. Closed‐form expressions for the parameter vector and the expected squared residual error are obtained by assuming that the observed covariate data are sampled from an equispaced design and that measurement error is generated from a symmetric probability distribution with a mean of zero. Notably, these results indicate that estimates of the amplitude parameters for a trigonometric regression model suffer from attenuation bias when covariate data are mis‐measured, and that estimates of the phase‐shift parameters are unbiased.

AbstractOften in surveys, key items are subject to measurement errors. Given just the data, it can be difficult to determine the extent and distribution of this error process and, hence, to obtain accurate inferences that involve the error-prone variables. In some settings, however, analysts have access to a data source on different individuals with high-quality measurements of the error-prone survey items. We present a data fusion framework for leveraging this information to improve inferences in the error-prone survey. The basic idea is to posit models about the rates at which individuals make errors, coupled with models for the values reported when errors are made. This can avoid the unrealistic assumption of conditional independence typically used in data fusion. We apply the approach on the reported values of educational attainments in the American Community Survey, using the National Survey of College Graduates as the high-quality data source. In doing so, we account for the sampling design used to select the National Survey of College Graduates. We also present a process for assessing the sensitivity of various analyses to different choices for the measurement error models. Supplemental material is available online.

International audience ; Proxy Means Testing (PMT) is a popular method to target the poor in developing countries. PMT usually relies on survey-based consumption data and assumes random measurement errors – an assumption that has been challenged by recent literature. Using a survey experiment conducted in Tanzania, this paper brings causal evidence on the impact of non-random errors on PMT performances. Results show that non-random errors bias the coefficients from PMT models, resulting in a 5 to 27 per cent reduction in PMT predictive performances. Moreover, non-random errors induce a 10 to 34 per cent increase in the incidence of targeting errors when poverty is defined in absolute terms. More reassuringly, impacts on the ranking of households are smaller and essentially non-significant. Taken together, these results indicate that PMT performances are quite vulnerable to non-random errors when the objective is to target absolutely poor households, but remain largely unaffected when the objective is to target a fixed share of the population.