The basic prerequisite for planning a plant or animal-breeding program is the total variability existing in the population and how much of it is caused by differences in the genetic makeup of the individuals. A quantitative measure of this genetic variability is provided by the genetic parameter `heritability'. The precise and accurate knowledge of heritability is very important as it expresses the reliability of the phenotypic values as a guide to the breeding value. Although a number of estimators of heritability are available in the literature, only very few have an exact sampling variance expression. Thus, the reliability, accuracy and trustworthiness of estimates of heritability need to be examined for different sample sizes. To overcome the difficulties arising from the mathematical complexities of the exact formulae for variance, it is desirable that the precision of these estimators is determined by using the analytical methods. Efron (1979 and 1982) has shown that the bootstrap method correctly estimates the variance of a sample median, a case where jackknife is known to fail. He showed that in many complex situations, where statistics are awkward to compute, they might be approximated by Monte-Carlo `re-sampling'. Further, it needs no prior assumption about the distribution of the observations as well as estimators. Singh et al. (2001) obtained an optimum size of the sample by the parent-offspring regression method for independent bootstrap master samples. Also, the optimum family size, structure and number of bootstrap replications were studied by half-sib method (Singh et al., 2003). As master samples are much concerned to get accurate estimates in bootstrap technique, an in-depth study is required to know the vital role of bootstrap master sample in the process. Thus, the present investigation is conducted by considering a single bootstrap master sample for a particular sample size to get the optimum number of bootstrap replications and sample size for estimating the standard error of heritability.
The important simulation model for regression or correlation as given by Ronningen (1974) has been used. To describe this model, the phenotypic values of a particular trait has to be assumed in both the parent and the offspring (or the relations) and the coefficient of inbreeding is assumed to be zero. |