In Response To 'Re: Re: Re: Strange NdSolve solution'
Thank you Bill very much, this helps a lot. I do notice that at these precisions, the command Abs[N[y]-z/.s[]] would give a very large discrepancy still. This maybe the result that at such large values of t, the solution to the differential equation becomes highly non-linear, hence difficulty in approximation, but I am not sure. I tried to increase both the accuracy goal and working precision but the error is still too far from zero. I wonder how to make the error very small. Thanks.