Linearization Variance Estimators for Survey Data with Missing Responses

Taylor linearization is a popular method of variance estimation for complex statistics such as ratio and regression estimators and logistic regression coefficient estimators. It is generally applicable to any sampling design that permits unbiased variance estimation for linear estimators, and it is computationally simpler than a resampling method such as the jackknife. However, it can lead to multiple variance estimators that are asymptotically design unbiased under repeated sampling. The choice among the variance estimators, therefore, requires other considerations such as (i) approximate unbiasedness for the model variance of the estimator under an assumed model, (ii) validity under a conditional repeated sampling framework. For example, in the context of simple random sampling and the ratio estimator, , of the population total , Royall and