查看更多>>摘要：For capturing information from hard-to-reach groups, a new sampling method was introduced in 1997 called respondent driven sampling (RDS). In conventional sampling, respondents are selected at random and sometimes by providing incentives to respondents for participation. However, in RDS, researchers select a small group called seeds, and after collecting information from them, researchers provide recruitment coupons to the seeds to recruit peers from their social networks. This process continues in waves. There are no standard formats for a coupon; it includes the serial number that links the recruit to their recruiter so that incentives can be given for the recruitment. The recruitment in RDS is controlled by the participants, based on incentives and chai-referrals with word of mouth as a major source of communication between successive waves. Though there are other approaches available to collect information from rare groups, RDS is more efficient when the target groups are geographically dispersed. For rare groups with socially unacceptable characteristics such as illicit substance users, this approach may not be suitable. For target groups that are not networked also RDS may not be effective. However, for target groups that are difficult to access but networked, this approach can work well.
查看更多>>摘要：Decreasing survey response rates are of great concern to researchers since it affects the sample selection process and the validity of inferences. Also, non-response rates in smaller samples may result in longer period surveys and increased survey costs. Non-contacts and refusals are two major components for evaluation of surveys. One of the first studies of this issue reported that from 1952 to 1979 refusals increased in two major face-to-face interview studies, though it did not address noncontacts. In a later stage study for the period from 1979 to 2003, it was found that both refusals and noncontacts increased, especially after 1985. Other studies have registered similar trends for overlapping as well as subsequent periods. Gallop poll surveys have a shown drop in response rates from 28% in 1997 to 9% in 2012 to 7% in 2017. Some efforts were undertaken to review data from different countries on response and noncontact rates. Differences were found across countries on refusal rates and noncontact rates, such as the former increased and thye latter decreased. In the United States, the trend on refusals was more on telephone surveys compared to face-to-face interviews.
Luke J. LarsenJoanna Fane LinebackBenjamin M. Reist
2页
查看更多>>摘要：Major survey authorities have the expertise to identify non-response reasons so that they can identify the correct respondents and convert the non-respondents to respondents. Yet, in the United States, non-response/refusal rates are increasing. This study explores the work of Harris-Kojetin and Tucker (Ref. 1) that investigated the macro-level factors of survey refusal in the wake of the present increase in refusal rates. In the referenced study, the authors used current population survey (CPS) data and a time-series regression approach to investigate the effects of economic and political influences such as unemployment rate, presidential approval rating, inflation rate and consumer sentiment score on the refusal rate. The hypothesis is that a negative attitude towards the government and a weak economy may decrease the participation rate. However, (Ref. 1) found that though the former influences refusals, the latter has no significant influence on reduction of participation, but may increase the participation. The original study pertained to the period 1960-1988. This current study replicates their findings and extends the study to the period 1960-2015. The original model was refined using available predictors and information about CPS design. This new study also hypothesizes the same relation between economic and political factors and survey participation.
David De ConinckCeline WuytsGeert LoosveldtLaurie Peeters...
2页
查看更多>>摘要：Many efforts are undertaken to change survey non-respondents to respondents by more attempts, tailored advance, reminder letters, incentives and bonus arrangements for interviewers. The main concern of non-response is its bias on the inference. However, studies show that correlation between non-response rates and non-response bias is not strong. Usually additional field work efforts are used to reduce the non-response rate. Reissuing initial non-respondents involves large resources and efforts. But this may not produce the intended reduction of non-response bias. To overcome this problem, a recently adaptive survey design was tested. Rsults show that implementation of these designs by putting differential field efforts to different groups can substantially reduce non-response bias. Since resources are limited, it is necessary to have a detailed analysis of fieldwork allocation. This article investigates the common practice of reissuing the groups to interviewers who have achieved higher response rates in the initial fieldwork phase and whether such reissuing reduces non-response bias.
Floyd J. FowlerPhilip BrennerAnthony M. RomanJ. Lee Hargraves...
3页
查看更多>>摘要：If the estimates from a sample survey are to be indicative of the population, then each of the members of the population should have an equal chance of getting selected. Then when the sample is selected, the required information should be collected from a high percentage of the sample. The gold standard followed in the United States is to draw a probability sample of households and send interviewers to conduct surveys. This approach has been in practice since 1940, and is still continued in important surveys such as the current population survey, national crime victimization survey and national health interview survey. However, to reduce survey costs, telephone surveys have been used by random- digit dialing to replace household visits. In the 1980s and 1990s, this approach has been effective as in-person visits since most households had phones. In the last twenty years, landline phones have became archaic due to the extensive use of mobile phones, making telephone surveys ineffective. This leads to a need for researchers to look for alternate methods including online surveys, email surveys and surveys using phone to provide answers to an automated interviewer such as interactive voice recording (IVR). All of these approaches can be used in various combinations in mixed-mode surveys. This study examines the effects of non-response on two types of surveys: interviewers calling telephone numbers matching addresses or using an IVR. Telephone surveys may result in more refusals as it takes time to answer the whole survey and for IVR, respondents may not call the number provided. Earlier studies have shown that the effect of non-response of various surveys on estimates are inconsistent. In some cases, the estimates are found to be good compared to in-person surveys, while others are not. The primary objective of this study is to assess the effect of non-response using telephone and IVR for major government surveys.
查看更多>>摘要：A scientific method for sample selection is the probability sample. All individuals should have a non-zero probability of selection and all of them should be known. This facilitates precise and unbiased estimates of population characteristics while quantifying the precision as well by means of confidence intervals or margin of error. Though this is the ideal situation, there are many factors that lead to poor estimates, the most conspicuous one is non-response bias. One major impact of non-response is that the sample size gets reduced. This can be overcome by increasing the initial sample size. The more serious problem is the bias potential on the population estimates. This can happen because of over- or under-representation of certain groups due to non-response. Often these differences in a population are due to ethnic or socioeconomic factors. To avoid this, the amount of non-response should be kept to a minimum. Estimating response probabilities depend on the model that is used. The most common model used is a logit model and this study compares this with the simple linear model. Estimation of response probabilities need individual values of auxiliary variables for both respondents and non-respondents which may not be available. This article explores approaches that do not depend on such preconditions such as using weights that can be used to estimate the response probabilities. These estimated response probabilities can be used for analysis of non-response, correction for non-response and as a measure of representativity indicator. This study also compares the response probability estimation using a logit model that is the benchmark for response probabilities from the linear model, transforming weights into probabilities using regression estimation and transforming weights by raking ratio estimation into estimated response probabilities.
查看更多>>摘要：In many surveys, it has been noticed that non-response rates are on the increase and it is necessary to understand the impact on the estimation of population parameters, or the non-response bias. The covariance between variables that affect the probability of responding and a given survey variable is of interest to researchers. Since the non-response bias is variable dependent, detecting its presence and assessing its impact are major challenges to researchers. To overcome this challenge, many indicators are proposed for the assessment of non-response bias. The most popular among them are the representativity indicator (R-indicator) along with the partial R-indicator. The R-indicator offers an approach for summarizing the extent to which the respondents in a probability-based sample survey represent all sample units selected and their non-response bias. Partial R-indicators are used to develop survey designs to identify specific subgroups for targeted interventions. The usefulness of R-indicators depends on the availability of auxiliary data for their estimation, which may not always be availabe. The auxiliary data can be used for adjusting the non-response bias. The effect of such adjustments on R-indicators is also a matter of interest. This study investigates the potential limitations of R-indicators for a case study from the Swiss European Social Survey, which in round five, included a non-response followup survey.
查看更多>>摘要：Though the general linear mixed model for data analysis is highly popular, the power and sample size methods are applicable only to special cases and in other situations, only rough approximations are used. While no software packages are available for this, accurate methods and software for power and sample size are available for linear multivariate models. Many applications in health sciences use special cases corresponding to multivariate models. It is necessary to know the conditions under which a mixed model can use power and sample size methods for a multivariate linear model. There are types of general linear mixed models including longitudinal and cluster designs that can use power and sample size methods and existing software. Though all multivariate models can be stated as mixed models, the converse is not true. Any mixed model can be converted into a multivariate model and vice versa meaning the process is reversible. So for reversible models, the power and sample size can be exactly computed. The objective of this study is to identify if reversible models are practical. The test statistic in many mixed models corresponds to that of multivariate test statistic for which power and sample size can be derived. The method is to use Wald test of fixed effects in the general linear model which has power equivalent to that of a general linear hypothesis test in the general multivariate model, and the power-equivalent hypothesis is reversible.
查看更多>>摘要：Lean Six Sigma (LSS) implementation is associated specific tools and it is necessary to understand the pros and cons of using various tools for an organization's improvement needs. Previous studies have highlighted the tools which are most popular. This study looks at LSS tools from a three-dimensional perspective- implementation, necessity and sufficiency. By LSS tools it is implied that the tools used in define-measure-analyze-improve-control steps in LSS projects. The details of the tools normally used for LSS projects are described in detail. A total of 106 respondents who are LSS team leaders are asked to rate the level of implementation, necessity and sufficiency of 68 LSS tools. The data was analyzed to examine how LSS tools are scattered, classified and discriminated to understand their relative importance. This analysis is done using Ward's hierarchical clustering and linear discriminant analysis within the three-dimensional space.
查看更多>>摘要：Lean Six Sigma (LSS) application in public organizations such as local, state or federal goverments is a difficult task since the organizational structure well as its administration are complex. Since profit making is not the primary motive, these organizations are interested in spending within budget rather than reducing cost. However, implementation of LSS can result in waste reduction as a part of Six Sigma. The define-measure-analyze-improve-control cycle provides a scientific approach to Lean projects and is are applicable to both small and large projects. Implementation of LSS requires the presence of critical success factors (CFS) in the organization and customization of public service organizations. This can be done by pursuit of main CFS identified in the literature before implementation of LSS. This article examines synergies between relevant CFS of LSS and public values in the public sector. This will be useful for public managers to successfully adopt LSS in their organizations.
NSTL