Abstract: For fifteen years, part of the work of our research team has been focused on the study of parental links between individuals living hundreds or thousands of years ago, whose remains have been found in single graves or large funerary complexes. These studies have been undertaken using methods developed by forensic genetics to identify individuals, mainly based on the genotyping of autosomal STR (Short Tandem Repeats). Issues arose from this work, namely the limits of studying small numbers of subjects, originating from groups of finite sizes where kinships cannot be inferred a priori and for which reference allelic frequencies do not exist. Although ideal human populations are rare when undertaking such studies, the Yakuts of Eastern Siberia constitute a very advantageous model, with large numbers of small pastoral communities and well-preserved archaeological material. The study of kinship in the ancient Yakuts allowed us to highlight the difficulties in analysing genetic data from small ancient human groups and to develop a strategy to improve the accuracy of statistical computations. This work describes this strategy and possible solutions to the study of populations outside of the frame of reference of global meta-populations, due either to isolation, remoteness or antiquity.