The number of shared species is a basic biodiversity index for characterizing the species richness and similarity of two ecological communities. Traditionally, most ecologists estimate the number of shared species by using the naïve estimator, which is simply the number of observed shared species in the sample of data. The resulting naïve estimate generally displays a negative bias due to some of shared species may be absent in the sample. In this study, we develop a probability model on the observed frequency of the sample data. Thus, we estimate the model parameters as well as the number of shared species via a conditional likelihood approach. Simulation studies based on some artificial populations and a census dataset, consisted of two real forest plots, were carried out to evaluate the performance of the proposed method.