Treatment of weights adjustments in the presence of unit non-response for small domains: a simulation study

Abstract

The information that the respondent does not provide is known as non-response. Small domains have a chance of not responding to units. Domain estimation becomes very interesting in such a circumstance. These little domains are typically quite hard to come by. The government does, however, intend to offer the facilities that are needed in these areas. As a result, the supportive variable may be beneficial and enhance the data on the intended domain. The supporting factor is really beneficial. We plot and the positively skewed nature of the variables can be used to support the small domains. Therefore, the type of supportive variables can influence the model that is selected. We examined calibration based power estimation. For two distance functions, minimum chi-square and Hellinger, a simulation analysis has been presented in terms of the absolute relative bias and simulated relative standard error. The result demonstrates for all intended domains, investigated calibration estimate in the presence of unit non-response is more efficient to ratio, exponential and generalized regression estimates. Additionally, it has been observed that the power based estimators using the Hellinger estimate are inferior to the power-based estimate with chi-square distance in both situations with and without two phase sampling.