Analysis methods: Missing data, error control

 

The primary purpose of this paper is to draw attention to the problems associated with response error in surveys and to suggest, by example, a variety of techniques for managing response error within ...

The article discusses the problem of missing data, and suggests a number of ways in which it can be addressed, including a thorough definition of the nature, type and frequency of the missing informat ...

Examines correlation, asking whether it necessarily yields the ‘best fit’ and whether this concept in necessarily objective. Similarly, does the correlation procedure have any intrinsic power to predi ...

Introduces a new conceptual framework, the customer targeting funnel model, to address customer access when there is limited information in the relevant customer database. Means of identifying those m ...

The USA’s National Do Not Call Registry (DNC Registry) had more than 91 million numbers registered by early 2005. The implications of this for survey response rates are explored in this paper, and amo ...

A dual frame survey of landline and cell phone (mobile phone) numbers was undertaken to evaluate the feasibility of using cell phone numbers in USA random digit dialling surveys. Some sources of nonre ...

Concerns about declining response rates in random digit dialling (RRD) telephone surveys were investigated by repeating a 1997 methodological survey. The findings again suggested that the level of uni ...

The author examined the hypothesis that respondents recruited by persuasive efforts may provide data filled with measurement error. Findings suggested, amongst other things, that the relationship betw ...

The link between nonresponse rates and nonresponse bias arises from the presence of a covariance between response propensity and the survey variables of interest. The paper suggests that the key commo ...

The authors report on a project to examine the potential effects of non-response via analyses that (a) investigate the linkages between community-level (USA zip-code) variables and nonresponse and (b) ...

The response rate for the American Time Use Survey (ATUS) has been below 60% for the first two years of its existence, raising questions about whether the results can be generalised to the target popu ...

The author explores whether probability sampling remains relevant when nonresponse is high and increasing and when there are cases where the linkage between nonresponse and nonresponse bias is absent. ...

Choice decisions involving groups of individuals are problematic. Individual preferences are easily measured, but joint preferences are expensive to obtain. The author proposes a sub-sampling approach ...

Multiple imputation is a well-established technique for analysing data sets where there is missing data. If the imputation model is correct, the resulting estimates are consistent. Alternative methods ...

Against a background of falling response rates, the development of computer-assisted methods for data collection has provided survey researchers with tools to capture a variety of process data (‘parad ...

‘Product recommendation systems’ are increasingly important to the network economy, leveraging customers’ prior product ratings to generate subsequent suggestions. The authors discuss increasing data ...

An integral part of the 2001 UK census was the creation of a transparent census database that adjusted for under-enumeration. This paper describes the methodology used, based on a controlled donor imp ...

Missing data occurs in most research surveys. The multiple imputation approach which is described in this paper can help to address this problem and is considered by some to be the gold standard for d ...

Latent class analysis (LCA) is a statistical tool for evaluating the error in categorical data when two or more repeated measurements of the same survey are available. The paper illustrates an applica ...

The authors explore the construction of more precise estimates of a collection of population means, using information about a related variable in the context of repeated sample surveys. Poststratifica ...

To cope with item and unit non-response, and to obtain adjusted means and proportions, a response propensity score is often used. The paper discusses the use of latent variable models with observed co ...

Explores ways in which missing data (which can otherwise result in loss of efficiency, and bias) can be imputed. Applies a method of multiple imputation to deal with missing data on alcohol consumptio ...

Explores the situation where observations are missing at random from a simple random sample drawn from a finite population. In some situations it can be of interest to create a full set of sample valu ...

Techniques to fill gaps in census data, using coverage measurement surveys and demographic information, are suggested. Analysis has suggested that unadjsted US Census counts are serious flawed for gro ...

This article presents a method that can be used to analyse a series of independent cross-sectional surveys where data may be missing through non-response or a variable methodological approach. The aut ...

As participants in conversation, we regard remarks such as 'I'm not sure' or 'I think' as indications that the speaker is unsure of the information he or she is providing. Subsequent discussions or a ...