Some estimation and bias issues in business surveys
Some estimation and bias issues in business surveys
The introduction of new businesses on a business frame is subject to reporting delays, that is, there are delays between the time when they have started trading and the time when they appear on the frame. Reporting delays cause undercoverage. The thesis provides methodology to predict the undercoverage, exploiting some links to AIDS research and actuarial science.
Another issue addressed is the bias that will arise if the knowledge of overcoverage gained in sample surveys is mistreated. It is usually discovered in the data collection phase of a survey that some units in the sample are ineligible even if the frame information has indicated otherwise. This information may be fed back to the frame and used in subsequent surveys, thereby making forthcoming samples more efficient by avoiding sampling ineligible units. The thesis investigates what effect on survey estimation the process of feeding back information on ineligibility may have, and derives an expression for the design-bias that can occur as a result of feeding back.
Although asymptotically design-unbiased, widely used GREG estimators may produce bad estimates. The thesis examines the behaviour of GREG estimators when the underlying model is misspecified. A diagnostic for whether a GREG estimate is reasonable is discussed. A common justification for the use of GREG estimators is that, being asymptotically designed unbiased, they are relatively robust to model choice. However, this work shows that the property of being asymptotically design unbiased is not a substitute for a careful model specification search.
The thesis raises the question of what the desirable properties of an estimator are and explores several point estimators in a simulation study. Special consideration is given to how prone an estimator is to produce large errors. This property is particularly important in official statistics where the publication of bad estimates may sometimes lead to great losses for society and may also be detrimental to the reputation of the producer.
University of Southampton
Hedlin, Dan Erik
4fcba8fd-9d9f-4225-9ede-c743eee6c0b4
2003
Hedlin, Dan Erik
4fcba8fd-9d9f-4225-9ede-c743eee6c0b4
Hedlin, Dan Erik
(2003)
Some estimation and bias issues in business surveys.
University of Southampton, Doctoral Thesis.
Record type:
Thesis
(Doctoral)
Abstract
The introduction of new businesses on a business frame is subject to reporting delays, that is, there are delays between the time when they have started trading and the time when they appear on the frame. Reporting delays cause undercoverage. The thesis provides methodology to predict the undercoverage, exploiting some links to AIDS research and actuarial science.
Another issue addressed is the bias that will arise if the knowledge of overcoverage gained in sample surveys is mistreated. It is usually discovered in the data collection phase of a survey that some units in the sample are ineligible even if the frame information has indicated otherwise. This information may be fed back to the frame and used in subsequent surveys, thereby making forthcoming samples more efficient by avoiding sampling ineligible units. The thesis investigates what effect on survey estimation the process of feeding back information on ineligibility may have, and derives an expression for the design-bias that can occur as a result of feeding back.
Although asymptotically design-unbiased, widely used GREG estimators may produce bad estimates. The thesis examines the behaviour of GREG estimators when the underlying model is misspecified. A diagnostic for whether a GREG estimate is reasonable is discussed. A common justification for the use of GREG estimators is that, being asymptotically designed unbiased, they are relatively robust to model choice. However, this work shows that the property of being asymptotically design unbiased is not a substitute for a careful model specification search.
The thesis raises the question of what the desirable properties of an estimator are and explores several point estimators in a simulation study. Special consideration is given to how prone an estimator is to produce large errors. This property is particularly important in official statistics where the publication of bad estimates may sometimes lead to great losses for society and may also be detrimental to the reputation of the producer.
Text
904800.pdf
- Version of Record
More information
Published date: 2003
Identifiers
Local EPrints ID: 464986
URI: http://eprints.soton.ac.uk/id/eprint/464986
PURE UUID: 1ac181d3-7b24-4c41-8d0d-4ea051b34eaf
Catalogue record
Date deposited: 05 Jul 2022 00:15
Last modified: 16 Mar 2024 19:52
Export record
Contributors
Author:
Dan Erik Hedlin
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics