The unitary PMNS matrix is the parametrization used for mixing of three active neutrinos. In case there are extra neutrinos which are heavy iso-singlets, the effective mixing matrix for the three active neutrinos will be non-unitary. As a result, the oscillation probabilities between the active neutrinos are modified from the probabilities obtained using the PMNS matrix. Long baseline neutrino experiments can probe the non-unitarity and constrain parameters of the model. We have analyzed the latest data from the NOvA and T2K experiments using both the unitary and non-unitary mixing scenarios. In this talk I will present results from our analysis and show that the tension between the oscillation data from these two experiments can be reduced when analyzing with non-unitary mixing of the neutrinos.