Journal of Graduate Studies - VOL - 41-2
Permanent URI for this collectionhttps://repository.neelain.edu.sd/handle/123456789/11808
Browse
Item Dealing with unit non-response in longitudinal sample surveys: a tailored weighting approach for subgroups(جامعة النيلين - مجلة الدراسات العليا, 2018-04-15) Husam Sadig; Majda MohamedLongitudinal surveys target large populations. Sample members from such populations are not usually homogeneous with respect to survey participation. Some sub-groups are more cooperative than others. This implies that the response propensity may be driven by different factors for different sub-groups in the same survey sample. Thus, non-response predictors in these sub-groups may differ from the general non-response predictors. However, this is overlooked in the traditional application of the logistic regression model which aims at predicting the probability of response and creating non-response weights for the sample as a whole. In most cases, the model is estimated using ‘generic’ socio-demographic variables and all units in the selected sample. Nonetheless, substantive analyses are often restricted to a sub-group of the sample and, therefore, if the general variables used to create the weights do not distinguish response from non-response in the sub-group in question, weighted estimates may be biased. This paper investigates whether weights are more effective if they are designed using variables correlated with the response propensity in the sub-group on which the analysis will be conducted. Using data from the British Household Panel Survey (BHPS), the findings demonstrate that, for some estimates, the tailored weights result in significantly different estimates than the standard weights.