Hi Rohit,
I suppose you are correlating observable particle_positions, i.e. a vector of the form (x1, y1, z1, x2, y2, z2 ... xN, yN, zN), then the output of the correlation square_distance_componentwise is (msd_x1, msd_y1, msd_z1 ... msd_xN, msd_yN, msd_zN), i.e. a 3N-dimensional vector, which gives you an individual msd per each particle and direction. The post-processing is left for the user, as the observable/correlation concept is more general and allows for more complex information than the msd. If you want msd from these numbers, you need to add up all the individual particle-direction msd's.
I am not sure how much of these features is presently accessible from python, sorry.