Ordinary Least Squares (OLS) Method is a good method to estimate regression parameters when there is no violation in classical assumptions, such as the existence of outlier. Outliers can lead to biased parameters estimator, therefore we need a method that can may not affected by the existence of outlier such as Minimum Covariance Determinant (MCD) and Least Median of Squares (LMS). However, the application of this method is less accurate when it is used for small data. To overcome this problem, it was aplicated bootstrap method in MCD and LMS to determine the comparison of bias in parameters which were produced by both methods in dealing outlier in small data. The used bootstrap method in this study was the residual bootstrap that works by resampling the residuals. By using 95% and 99% confidence level and 5%, 10%  and 15% outlier percentage, MCD-bootstrap and LMS-bootstrap give value of parameter estimators which were unbias for all percentage of outlier. We also found that the widht of range which produced by MCD-bootstrap method was shorter than LMS-bootstrap method produced. This indicates that MCD-bootstrap method was a better method than LMS-bootstrap method.