The objective of this research is to compare the estimation methods for nonignorable missing-data of the independent variables in binary logistic regression models with three independent variables. The estimation methods considered in study are Mean Imputation (MEAN), Median Imputation (MED), K-Nearest Neighbor (KNN) and Multiple Imputation (MI). Data of this research are simulated with three sample sizes of 70, 100 and 200. Three levels of missing proportion of data are 10%, 20% and 30% and three levels of nonignorable-missingness of data are ignorable, middle nonignorable and high nonignorable. Coefficients of three independent variables in simulation are set to be 0.5, 1 and 1.5, respectively. The comparison of each methods using the average mean square error (AMSE), the findings are as follows: i) MI method yield higher performance when coefficients of independent variables are low and small sample sizes, ii) MEAN method and MED method yield higher performance when coefficients of independent variables are high and large sample sizes. Iii) the AMSE increase when proportion of nonignorable-missingness of data increase, iv) the AMSE increase when coefficients of independent variables increases.