A bootstrap method to calculate the p -value of Fisher’s combination for a large number of weakly dependent p -values

In this paper, we propose an effective bootstrap algorithm to calculate the p-value of Fisher’s combination for a large number of weakly dependent p-values. The proposed bootstrap algorithm is based on the stationary bootstrap idea for dependent observations. The proposed bootstrap algorithm can control empirical sizes well and gain desirable empirical powers by extensive simulations. We also apply the bootstrap algorithm to a real data analysis about coronary artery disease.