Moran's I index is the usual tool to measure the intensity of the spatial autocorrelation in genetic markers data. I statistics is asymptotically normally distributed and it may be evaluated as standard normal deviations (assumption-N, normality). However, for small numbers of populations (m<8), the Mantel' s randomness test (assumption-R) developed by Mantel (1967) should be applied. Thus, this study was done to evaluate the performance of both tests accordding to type I error rate sand their power. They were evaluated via Monte Carlo simulation, in which, the situations of average allelic frequencies, {p=0,1, p=0,25 and p=0,5} were analyzed under H0. Number for populations varying from {m= 5, 10, 25 and 50}were taken into account and for each population, the number of individuals in {n=1, 2, 5, 10 and 30} was varied as well. As regards to the alternative hypothesis (with spatial pattern), in addition to these same situations simulated in H0, the behavior of these criteria of tests was evaluated according to the variation of the amplitude in the average local allelic frequency in {A=0,1; 0,2; 0,5; 0,8 e 1,0}. Therefore, the performance of the test studied could be analyzed as the degree of variability of the average frequencies generated on a linear surface, related to the geographic space and by means of it's different slopes. The normal approximation was considered better withpopulations as combined with the weighing systems inverse of the distance and inverse of the distance squared in both levels of significance 1% and 5%. The same should not be applied associated with the nearest neighbor weighing. With , Mantel's test should be applied in any of the situations simulated.
Spatial autocorrelation; Monte Carlo simulation; Moran's I; allelic frequencies; Mantel's test
