Thank you for your answer, I hope I understand it now. I calculated the MSD for the same trajectory using different combinations of ncors and nskip, each time charmm gives a different estimate of diffusion coefficient, of the order E00 to E-06, and even negative values. I want to ask how reliable this estimation is, or maybe I should plot the data and do the linear fit myself. I attached such a plot of MSD vs time and the plot is not that linear. Do you think it looks OK? Should I fit only the first 10 ps? In any way I fit it, I can not reproduce the estimated diffusion coefficient from charmm output. I would be very grateful if you could help me on this matter.