gfComparison of two Methods for Estimating Variance Components of Response Error Model in Surveys

Survey implementation is one of the common methods to data collection. To determine quality of survey results, survey errors should be studied. Response error is an essential part of nonsampling errors and modeling of response error has many applications. One of the applications is to estimate precision of survey results. In this paper, two methods of variance components estimation of the response error model is studied and compared. In this purpose, estimators of variance components is computed by Maximum Likelihood and ANOVA methods. As an application of the presented maximum likelihood estimators, variance components is estimated for some data sets and optimization is applied to assess validity of the proposed estimators. These calculations show that nonlinear optimization confirms the presented maximum likelihood estimators, in all of the data sets. Furthermore, Mean Squares Error criterion is applied to compare the presented maximum likelihood estimators with ANOVA estimators of variance components of the response error model. In the other words, variance and bias of the estimators is calculated to compare two methods of estimation.