Isn't there a problem for saving the coefficients of NLS estimations in a loop? No problem at all for OLS, but couldn't do the same with NLS