To compare 2 methods of variance components for factorial crossed classification design; the bootstrapping method and the classical method. Monte Carlo simulation is done through S-plus 2000 code. It is simulated under situation due to the distribution of random errors. When the distribution of random error is normal distribution, the simulation is specified at a=b=2, n=3,5 and 7;at a=b=3, n=3,5 and 7; and at a=b=4, n=3,5 and 7, the coefficient of variation (CV) is specified at 10%, 30% and 50% respectively. When the distribution of random error is contaminated normal distribution, the simulation is specified at a=b=2, n=3,5 and 7; at a=b=3, n=3,5 and 7; and at a=b=4, n=3,5 and 7, the variance of random eror sigma2 is specified at 100, the percent of contamination is specified at 5%, 10% and 25% while the scale factor for contaminated variance of the errors is generated at scale factor 3 and 10. The average of Euclidean distance between the vector of variance component estimates and the vector of true is ameasure for comparison between both methods. The results of this study show that when the random errors have normal distribution, the number of levels for both factor A and factor B and the number of replication are low (a=2,b=2, and n=3 or 5), the point estimates for variance components using the bootstrapping method, provide shorter averaged distance than the one from the classicical method. When the number of levels for both factor A and factor B, and the number of replication are high (a=4,b=4,and n=5 or 7), the distance from the bootstrapping method is bigger than the one from the classical method. When the distribution of random errors is contaminated normal distribution regardless the percent of contamination and the scale factor, the distance using the bootstrapping method is shorter than the one from classical method in almost of all case (98%).
