|| Checking for direct PDF access through Ovid
The sampling scheme for a core collection from the base collection of Yunnan rice landraces was studied, using 6,121 accessions catalogued in the national genebank and 31 characters including taxonomic characters, morphological characters and agronomic characters. Sampling scheme was studied at three levels, i.e., grouping principle, sampling proportion within group and sampling method from each group. Grouping principles used were Dingying's taxonomic system, Cheng-Wang taxonomic system, Yunnan rice ecological zone, Yunnan administrative district, single character and completely random sampling without grouping. Sampling proportions in each of the groups were square root, logarithm, index of genetic diversity and fixed proportion. Clustered and random sampling were carried out in each group. Five parameters, (index of genetic diversity, variance of phenotypic value, variance of phenotypic frequency, coefficient of variation and ratio of phenotypic retained), in primary core collection were used to evaluate the validity of sampling scheme. The better sampling strategies were grouping principles based on the two taxonomic systems, decisions on sampling proportion in each group based on square root or logarithm, sampling method within group clustered. 16%, 10% and 5% overall sampling ratio were used to establish primary core collections. The results showed that a 5∽10% sampling ratio was suitable for a large collection.