ABSTRACT
In different statistical studies, a subset of data maybe missing for some study subjects either by design or happenstance, such data are called missing variables. Ignoring such data causes bias in the results; therefore presenting statistical methods for analyzing such data are necessary. One of the most common techniques used in linear regression analysis with missing covariates is a Weighted Estimating Equation [WEE]. In this method, the observe probability of missing data are computed using the logistic regression, then the inverse probability of these data are input into the score statistics equation and finally the equation is solved using the EM algorithm and the regression parameters are estimated. The advantage of this method is that the distributions of the missing data need not to be correctly specified. In the present study, the above method was compared to Maximum Likelihood [ML] by using an applied example. Considering the covariates missing at random [MAR], the WEE method is more efficient than the other statistical methods. Regarding the advantages of WEE, this method is applicable when the distributions of covariates are not normal