Statistics for Storage and retrieval of large data sets: Dimensionality reduction and nearest neighbour search