For the one-way classification random model with unbalanced data, I compare five estimators of σ 2/a and σ 2/e the among and within-treatments variance components: Analysis of Variance (ANOVA), Minimum Norm Quadratic Unbiased Estimators (MINQUE) and the rest are non-negative estimators: the Positive Semi Definite Minimum Norm Quadratic Minimum Biased Estimators (PSD-MINOMBE), the Closest to MINQUE (CMINQUE) and the Closest to ANOVA (CANOVA) estimators. The estimators are compared through their biases and MSE's, which are estimated by the Monte Carlo simulation technique.