An iterative estimating procedure for probit-type nonresponse models in surveys with call backs

Test (Impact Factor: 0.98). 02/2000; 9(1):233-253. DOI: 10.1007/BF02595860
Source: RePEc


This work attempts to treat the negatives to respond in sample plans when several tries or call backs in the capture of individual
data are assumed. We also maintain the assumption that the respondents supply all the variables of interest when they are
captured although the retries are kept on, even after previous captures, for a predetermined number of tries,r, fixed only for estimating purposes. Supposing that the different retries or call backs are exerted with different capture
intensities, the response probabilities, which may vary from one individual to another, are searched by probit models whose
parameters are estimated using conditional likelihoods evaluated on the respondents only (other models, derived from error
distributions different from normal, could also be possible by approximating numerical techniques quite similar to the ones
proposed here). We present a numerical estimating algorithm, quite easy to implement, which may be used when the recorded
information about data captures includes at least marginal results. Finally, we include some encouraging empirical simulations
whose purpose is centred in testing and evaluating the practical performance of the procedure.

1 Read
  • Source
    • "Concerns about endogeneity of the number of contact attempts with the outcome of survey participation and " censoring " for cases that are never contacted or interviewed have led to the use of discrete time hazard models that change the outcome to the conditional probability of an interview on a given call, given no contact or participation on prior calls (Durrant and Steele 2009; Groves and Heeringa 2006; Kennickell 1999; Olson and Groves 2009). Other probability-based models have been used to estimate response probabilities at each call attempt as a function of respondent characteristics, sometimes permitting a " hard core " nonresponding group (Alho 1990; Anido and Valdés 2000; Colombo 1992; Drew and Fuller 1980; Potthoff, Manton, and Woodbury 1993; Wood, White, and Hotopf 2006). A recent expansion of the callback models uses latent class models, characteristics of respondents and nonrespondents, and reports from the survey to create weights based on the level of effort exerted to the case (Biemer 2009; Biemer and Wang 2007; Biemer and Link 2006). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Survey researchers and practitioners use nonresponse adjustment weights to mitigate the effects of survey nonresponse on sample estimates. One challenge in creating these weights is finding useful auxiliary data that predict both the probability of participating in the survey and the survey variables of interest. This article reviews the use of paradata for nonresponse adjustment. Five different types of paradata are considered: neighborhood observations, observations of the sampled housing unit, observations of persons in the sampled housing unit, call records, and observations about the interviewer-householder interaction. Empirical evidence about the predictive value of these paradata for predicting both participation and survey variables is examined. Challenges of using these paradata are also identified, along with outstanding issues and opportunities related to the use of paradata for nonresponse adjustment.
    The Annals of the American Academy of Political and Social Science 01/2013; 645(1-1):142-170. DOI:10.1177/0002716212459475 · 1.01 Impact Factor
  • Source
    • "Kuk et al.(2001) suggested imputation and prediction approaches in estimating finite population quantities. Approaches have also been suggested by Potthoff et al. (1993), Anido and Valdes (2000) and Copas and Farewell (1998). In this paper we show how to estimate and adjust for non-ignorable non-response under two assumptions. "
    [Show abstract] [Hide abstract]
    ABSTRACT: We present a general method of adjustment for non-ignorable non-response in studies where one or more further attempts are made to contact initial non-responders. A logistic regression model relates the probability of response at each contact attempt to covariates and outcomes of interest. We assume that the effect of these covariates and outcomes on the probability of response is the same at all contact attempts. Knowledge of the number of contact attempts enables estimation of the model by using only information from the respondents and the number of non-responders. Three approaches for fitting the response models and estimating parameters of substantive interest and their standard errors are compared: a modified conditional likelihood method in which the fitted inverse probabilities of response are used in weighted analyses for the outcomes of interest, an EM procedure with the Louis formula and a Bayesian approach using Markov chain Monte Carlo methods. We further propose the creation of several sets of weights to incorporate uncertainty in the probability weights in subsequent analyses. Our methods are applied as a sensitivity analysis to a postal survey of symptoms in Persian Gulf War veterans and other servicemen. Copyright 2006 Royal Statistical Society.
    Journal of the Royal Statistical Society Series A (Statistics in Society) 07/2006; 169(3):525-542. DOI:10.1111/j.1467-985X.2006.00405.x · 1.64 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The paper considers the use of level-of-effort (LOE) paradata to model the non-response mechanism in surveys and to adjust for non-response bias, particularly bias that is missing not at random or non-ignorable. Our approach is based on an unconditional maximum likelihood estimation (call-back) model that adapts and extends the prior work to handle the complexities that are encountered for large-scale field surveys. A test of the 'missingness at random' assumption is also proposed that can be applied to essentially any survey when LOE data are available. The non-response adjustment and the test for missingness at random are then applied and evaluated for a large-scale field survey—the US National Survey on Drug Use and Health. Although evidence on non-ignorable non-response bias was found for this survey, the call-back model could not remove it. One likely explanation of this result is error in the LOE data. This possibility is explored and supported by a field investigation and simulation study informed by data obtained on LOE errors.
    Journal of the Royal Statistical Society Series A (Statistics in Society) 01/2013; 176(1):147-168. DOI:10.2307/23355181 · 1.64 Impact Factor