In order to determine the Lotka parameters for a bibliography, one usually uses the complete data set. In this paper it is shown that it is possible to use only a fraction of the original data, namely by sampling randomly. However, sampling can be done either by source, i.e. selecting a fraction of the authors, or by item, i.e. selecting a fraction of the publications. It is shown here both by experiments, using computer simulations, and by mathematical approach, that only sampling by source is allowed for the mentioned purpose. Item samples give a completely disturbed idea about the Lotka law for the bibliography. From source sample size equal to 10% onwards, one gets good results. For the calculation of the Lotka exponent, a known, simple and fitting method is used and refined.