Recently, a positive spin Hall angle (SHA) of 0.021 was observed experimentally in nonmagnetic CuIr alloys [Niimi et al, Phys. Rev. Lett. 106, 126601 (2011)] and attributed predominantly to an extrinsic skew scattering mechanism, while a negative SHA was obtained from ab initio calculations [Fedorov et al, Phys. Rev. B 88, 085116 (2013)], using consistent definitions of the SHA. We reconsider the SHA in CuIr alloys, with the effects of the local electron correlation U in 5d orbitals of Ir impurities, included by the quantum Monte Carlo method. We found that the SHA is negative if we ignore such local electron correlation, but becomes positive once U approaches a realistic value. This may open up a way to control the sign of the SHA by manipulating the occupation number of impurities.