Abstract

CARLINI-GARCIA, Luciana Aparecida; VENCOVSKY, Roland  and  COELHO, Alexandre Siqueira Guedes. Bootstrap method applied over resampling levels in the estimation of genetic parameters of populations. Sci. agric. [online]. 2001, vol.58, n.4, pp. 785-793. ISSN 0103-9016.  http://dx.doi.org/10.1590/S0103-90162001000400021.

Genetic structure and reproductive system of natural populations have often been studied using molecular markers and isozymes. Specific genetic parameters are therefore being estimated for this purpose. There is little information about errors associated with these estimates due to different possible sampling levels. Some real datasets were used in this study and bootstrapping was performed over loci, individuals, populations and both individuals and populations simultaneously. These different resampling strategies produced associated errors of the estimates and their empiric distributions and confidence intervals for the following parameters: total fixation index (F ), fixation index within populations (f ), diversity between populations (q ) and apparent outcrossing rate (t_a). Generally, smaller errors were associated with . The only empirical distribution of estimates approaching normality were for and when individuals were envolved in the resampling process. Bootstrap samples of variable sizes were used to obtain the number of loci, individuals and populations necessary to achieve a given precision of F, f and q. Generally, the magnitude of errors of the estimates in most datasets indicated that sample sizes currently used in research involving natural populations were suitable only for the estimation of q. The largest errors of and , were associated to sampling errors due to loci. Therefore, future studies should use larger numbers of loci. Sampling errors due to population was also important and should not be neglected.

Keywords : population structure; population genetics; estimation.