Neutrino oscillations are studied in T2K by comparing the observed data at the far detector to the predictions for different hypotheses to make statistical inferences. The predictions are made using a model of the experiment, constructed from a set of models and simulations tuned to data to predict the neutrino flux reaching the detectors, and describe neutrino interactions as well as the detectors response. Using data corresponding to an exposure of 1.49$\times 10^{21}$ proton on targets (POT) in neutrino mode, and 1.63 $\times 10^{21}$ POT in anti-neutrino mode, T2K excluded conservation of CP symmetry in neutrino oscillation at the 2$\sigma$ level. An excess of observed events compared to the predictions in one of the samples has a strong contribution to this result, and it was found that the p-value to observe such an excess in 1 of the 5 analysis samples was between 5.8\% and 11.3\% depending on the values of the oscillation parameters assumed.