Abstract
The most often used non-uniform distribution in applications involving simulations is the Gaussian distribution, popularly referred to as the Normal distribution. This study examines performances of different Gaussian variate generators over small, moderate and large samples with respect to statistical accuracy of the parameter estimates produced and computational complexity. Results showed that in statistical accuracy, the Marsaglia-Bray’s algorithm performed best in small, moderate and large samples with a maximum absolute error of 1.08; while in computational efficiency, the Box-Muller’s algorithm took a bit longer to compute the normal variate Z compared to the other algorithms considered.