LAUSR

In this paper, we consider an exact test of goodness of fit for binomial distribution in sparse data situation. A conventional way is viewing this problem as an independence test problem of a two-way contingency table. We propose an approach to promote the efficiency of the Diaconis–Sturmfels (DS) algorithm when n (sample size) is much larger than m [the first parameter of a binomial distribution B(m, p)] through representing the data and then utilizing minimal Markov bases of the corresponding multinomial model. Simulation results and real data analysis indicate that our method makes the DS algorithm computationally faster.